University of Southern California Dissertations and Theses

Adolescent stress and styles of coping: Predictors and moderators of psychological distress

(USC Thesis Other)

Adolescent stress and styles of coping: Predictors and moderators of psychological distress

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content INFORMATION TO USERS This m anuscript has been reproduced from the microfilm master. UM I films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type o f com puter printer. T h e quality o f th is rep ro d u ctio n is d ependent upon the quality o f th e copy su b m itted . Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and im proper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a com plete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion. Oversize m aterials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand com er and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back o f the book. Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6” x 9” black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UM I directly to order. UMI A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor MI 48106-1346 USA 313/761-4700 800/521-0600 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ADOLESCENT STRESS AND STYLES OF COPING: PREDICTORS AND MODERATORS O F PSYCHOLOGICAL DISTRESS by Benjamin Isaac Broder A Dissertation Presented to the FACULTY OF TH E GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (Preventive M edicine — Health Behavior) December 1995 Copyright 1995 Benjamin Isaac Broder Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. ÜMI Number: 9630742 Copyright 1995 by Broder, Benjamin Isaac All rights reserved. UMI Microform 9630742 Copyright 1996, by UMI Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. UMI 300 North Zeeb Road Ann Arbor, MI 48103 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. UNIVERSITY OF SOUTHERN CALIFORNIA THE GRADUATE SCHOOL UNIVERSITY PARK LOS ANGELES, CALIFORNIA 90007 This dissertation, written by Benjamin Isaac Broder under the direction o f his Dissertation Committee, and approved by all its members, has been presented to and accepted by The Graduate School, in partial fulfillment o f requirements fo r the degree of DOCTOR OF PHILOSOPHY C. / Dean o f the Graduate School Finto November 14, 1995______ DISSERTATION COMMITTEE Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. D e d i c a t i o n This dissertation is dedicated to my wife, Kimiko. Without her support and encouragement it would have been an impossible task. She encouraged me when I was discouraged, prodded me when I lacked motivation, and helped me always. It is also dedicated to my daughter, Lauren, whose smile lights up the room, makes my worries fall away and puts the rest of my life in perspective. It is dedicated to my grandfather, Leonard TYishnet, M .D ., who served as an example of a caring physician who balanced family and career, clinical practice and writing. His death was a great loss for our entire family, but he lives on in the memories of all who knew him. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill A c k n o w l e d g e m e n t s I wish to express my gratitude to the many people who contributed to the completion of this project. I am deeply indebted to Dr. M ary Ann Pentz, who served as my advisor and the Chair of my dissertation committee. M ary Ann has been an inspirational mentor throughout my association with her. She has always been accessible when I needed her advice, even when the logistics of our interaction were daunting. She always believed in me and pushed me to achieve at a level of excellence that I may not have reached for on my own. My collaborations with her have taught me more about research excellence than any other graduate educational experience. I am also indebted to Dr. Chih-Ping Chou, without whose analytic insight I would not have been able to complete this project. W ithout exception he made himself available to me for advice and consultation. I hope to emulate his example as a patient and responsive teacher during my career. I am very grateful to Dr. Victor Henderson, who agreed to serve on my dissertation committee although he was already extremely busy. His perspective as a clinician-researcher has been invaluable and his advice during the planning stages of this dissertation was invaluable. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IV I wish to express my gratitude to Dr. M alcolm Pike, for taking the time to step outside his area of expertise and serve on my dissertation committee. His belief in the importance of this work has been very important to me. I am thankful that I have had the opportunity to work with and study under Dr. C. Anderson Johnson. Andy has shown all of us at IPR what is possible in prevention research and how one can move from research interest to research interest during a career while remaining responsive and interested in the education of graduate students. I wish to thank the other members of the M idwestern Prevention Project team. Dr. Suzanne Montgomery, Dr. Luanne Rohrbach, and Dr. Carol Hodgson were enthusiastic and supportive of me throughout my time at IPR. Our collaboration on papers was invaluable in teaching me how to work as a research team. In particular, Luanne and Carol have, at separate times, been supportive and empathetic as I progressed from concept to completion of this dissertation. M artha Petrov has been a supportive friend during my years with the M idwestern Prevention Project, as well as providing outstanding administrative support. I am lucky to know her. Bobbie Searl has also been a friend and has followed in M artha’s footsteps in going above and beyond the call of duty in helping me with a variety of tasks during my work on this dissertation. Many other faculty members have had strong influences on me during my tim e at IPR. Dr. Jean Richardson has provided a stellar example of research Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. V excellence and sparked my interest in the area of stress and coping. Dr. Gary Marks, in my first graduate course at IPR, showed such a level of passion for both research and for teaching that I am inspired to bring such passion to my career. Dr. David M acKinnon was completely receptive to me when I first came to IPR and showed me how rewarding and exciting a professional collegial relationship can be. I will never foiget his willingness to sit and philosophize about prevention although considering my level of experience I must have brought very little to the conversation. Drs. Steve Sussman, Clyde D ent, and James Dwyer have also contributed to my education as a researcher by providing both educational opportunities in the classroom and personal examples outside the classroom of excellence in research and teaching. No acknowledgement would be complete without acknowledging the debt I owe my parents. M y m other and father have taught me to strive for excellence in whatever I desire. Their own lives have been outstanding examples to me of how loving your career, loving your spouse, and loving your children can improve each aspect of your life without subtracting from any of them. I have also learned much from my brother and sister, M ichael and Leah. M ike has always been my friend and confidante. We were thrown together by biology, but have stayed close because of the mutual support that we give each other. Two years behind me in school, for many years on a superficial level he followed in my footsteps, although he always brought a perspective to many Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. VI endeavors that added so much more than I ever could. In recent years I have been grateful to follow in his footsteps and have been able to rely on M ichael for much needed advice when applying to medical school, when beginning to work in clinical settings, when planning my wedding, and when raising my first child. His advice and support have been invaluable. Leah has taught me so much about life and standing up for and speaking out for what you believe in that I do not know where to begin to thank her. Whenever I feel a conflict between “doing things right” and “doing the right thing,” I can always think of what Leah would do (or ask her what she would do) to help me see the proper path. This dissertation was supported in part by NIH Grant DA03976. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. vil T a b l e o f C o n t e n t s Page D ed icatio n ............................................................................................................................. ii Acknow ledgem ents................................................................................................................... iii List of T a b le s ....................................................................................................................... xii List of Figures ....................................................................................................................... xiv A b s tr a c t....................................................................................................................................xvi In tro d u ctio n .......................................................................................................................... 1 Background and S ignificance........................................................................................... 3 Health Behavior and Health P ro m o tio n ......................................................... 3 Health Behavior ..................................................................................... 3 Health Promotion .................................................................................. 4 Adolescent Development and Stress ............................................................... 5 Stress and S tresso rs.............................................................................................. 9 Origins and Foundations ..................................................................... 9 Stressor T h e o rie s..................................................................................... 13 Stressor Measures .................................................................................. 16 Stressful Life Events and Adverse O u tco m es................................... 20 Coping .................................................................................................................... 22 Origins and Foundations ..................................................................... 22 Coping Theories ..................................................................................... 23 Function of C o p in g .................................................................. 24 M odes or Methods of C o p in g ............................................... 25 Coping M e a su re s..................................................................................... 27 Coping and Psychological Outcomes ............................................... 30 Psychological Symptoms and D is tr e s s ............................................................ 31 M ediator versus M oderator Distinction ......................................................... 34 Methods of T e s tin g .............................................................................................. 39 Conceptual M o d e ls.................................................................................. 40 Cross-Sectional M o d e ls ............................................................ 40 Distress Outcome Longitudinal M o d e l................................ 41 Fully Lagged Longitudinal M odel ...................................... 42 Change Scores Longitudinal M o d e l...................................... 43 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. vin M ethodological Issues ........................................................................................ 44 Implications for Health Promotion and Disease Prevention .................... 45 Study Hypotheses .............................................................................................................. 47 M ajor H y p o th eses................................................................................................. 48 Construct/M easurement Model Hypotheses ................................... 48 Hypothesis 1. Domains of S tr e s s o r s ................................... 48 Hypothesis 2. Domains of C o p i n g ...................................... 49 Structural H y p o th e s e s ............................................................................ 49 Hypothesis 3. Stressors and Psychological Distress . . . 49 Hypothesis 4. Avoidance Coping and Psychological D istre ss............................................................................ 50 Hypothesis 5. Avoidance Coping as a M e d i a t o r 50 Hypothesis 6. Approach Coping as a Stress Buffer . . . 51 M inor H y p o th eses................................................................................................. 52 Construct-Level Hypotheses ............................................................... 52 Hypothesis 7. Other Approach Coping E ff e c ts ................ 52 Sub-Group Hypotheses ........................................................................ 52 Hypothesis 8. Gender Hypotheses ...................................... 52 Hypothesis 8a. Gender and Psychological S y m p to m s......................................................... 52 Hypothesis 8b. Gender and C o p in g ...................... 52 Hypothesis 8c: Gender and Stressor/Coping/Symptoms R e la tio n sh ip s.................................................. 53 Hypothesis 9. Age-Related Hypotheses ............................. 53 Hypothesis 9a. Age and Stressful Life Events . . 53 Hypothesis 9b. Age and C o p in g ........................... 53 Hypothesis 9c. Age and Stressor/Coping/Symptoms R e la tio n sh ip s.................................................. 54 M e th o d s ................................................................................................................................. 54 Study D e sig n ........................................................................................................... 54 Data Collection P ro c e d u re s ............................................................................... 56 Sample .................................................................................................................... 58 M e a s u re s ................................................................................................................. 60 Statistical P o w e r.................................................................................................... 62 Data Analysis ........................................................................................................ 66 Data C lean in g ........................................................................................... 66 Sample C haracteristics............................................................................ 67 Scale C o n stru ctio n .................................................................................. 67 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. IX Tests of the M ain Hypotheses ............................................................ 68 Structural Equation M odeling O v e rv iew ............................. 68 M easurement M odel ............................................................... 72 Single Group Structural M o d e l ............................................ 74 M ultiple Group M o d e l ............................................................ 76 Regression M e th o d s ............................................................................... 78 R e s u lts ..................................................................................................................................... 79 Descriptive S ta tis tic s ............................................................................................ 79 Observed Variable C h a ra cteristics...................................................... 79 Factor Characteristics ............................................................................ 81 Sample Group Equivalence ............................................................................... 83 Approach Coping Group E quivalence................................................ 83 G ender and Grade Group Equivalence ............................................ 84 M easurement M odel V alid atio n ......................................................................... 87 Factor Analysis and Scale Construction ......................................... 87 M easurement Invariance Across Groups ......................................... 92 Confirmatory Factor A n a ly s is ................................................ 92 M odel M o d ificatio n .................................................................. 93 Regression M odels ............................................................................................... 95 Structural Equation M o d e l s .................................................................................. 100 In tro d u c tio n ..................................................................................................100 Distress Outcome M odel ........................................................................ 101 Separate M odels by G r o u p .........................................................102 M odel M o d ific a tio n ..................................................................... 102 Equality of Variable Means Across G r o u p s ......................... 106 Equality of Factor Loadings Across G ro u p s ......................... 107 Equality of Factor Correlations and Regression W e ig h ts...............................................................................107 Summary of Distress Outcome Longitudinal SEM M o d e l.................................................................................. 108 Fully Lagged M o d e l .................................................................................. I l l Separate M odels by G r o u p .........................................................I l l M odel M o d ificatio n ..................................................................... 112 Equality of Variable M eans Across G r o u p s ......................... 113 Equality of Factor Loadings Across G ro u p s......................... 113 Equality of Factor Correlations and Regression W e ig h ts...............................................................................114 Summary of Fully Lagged Longitudinal SEM M odel ..115 Change Scores M o d e l ...............................................................................117 Separate M odels by G ro u p .........................................................118 M odel M o d ificatio n ..................................................................... 118 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. X Equality of Variable Means Across G r o u p s ......................... 119 Equality of Factor Loadings Across G ro u p s ......................... 119 Equality of Factor Correlations and Regression W e ig h ts.............................................................................. 120 Summary of Change Scores Longitudinal SEM M o d e l..................................................................................121 1990 Cross-Sectional M o d e l ..................................................................123 Separate M odels by G r o u p .........................................................123 Model M o d ificatio n ..................................................................... 124 Equality of Variable Means Across G r o u p s ......................... 125 Equality of Factor Loadings Across G ro u p s ......................... 125 Equality of Factor Correlations and Regression W e ig h ts.............................................................................. 125 Summary of 1990 Cross-Sectional Longitudinal SEM M o d e l..................................................................................126 1991 Cross-Sectional M o d e l ..................................................................127 Separate M odels by G r o u p .........................................................127 Model M o d ificatio n .....................................................................128 Equality of Variable Means Across G r o u p s ......................... 129 Equality of Factor Loadings Across G ro u p s ......................... 129 Equality of Factor Correlations and Regression W e ig h ts.............................................................................. 129 Summary of 1991 Cross-Sectional Longitudinal SEM M o d e l..................................................................................130 Summary .................................................................................................... 131 D iscu ssio n .................................................................................................................................133 O v e rv ie w .................................................................................................................... 133 Stress and Coping Constructs (Measurement M o d e l ) ....................................133 Stress/Coping/Distress Relationships (Structural M o d e l ) ............................ 135 M ultiple Regression ................................................................................. 135 Structural Equation M o d e ls .....................................................................136 Implications for Prevention and Future Research D ire c tio n s......................138 L im ita tio n s.................................................................................................................149 R e lia b ility .................................................................................................... 149 Internal C onsistency.....................................................................149 Test-Retest R eliab ility .................................................................. 149 V alidity ...........................................................................................................150 M easurement Construct Validity ............................................150 Internal Validity of the Study .................................................. 151 External Validity of the Findings .........................................................153 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. XI R eferences................................................................................................................................ 165 Appendix ............................................................................................................................. 180 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Xll L i s t o f T a b l e s Page Table 1. Hypotheses and their Significant Model Implications............................... 55 Table 2. Sample C h aracteristics..................................................................................... 58 Table 3. Descriptive Statistics and Correlations of Factors by Adaptive Coping Group ........................................................................... 82 Table 4. Demographics by Approach Coping G r o u p ............................................... 83 Table 5. Demographics by Gender and G ra d e ............................................................ 84 Table 6. Stress, Coping, and Distress Factor Means by Gender and G r a d e .............................................................................................................. 85 Table 7. Measured Variable Means by G r a d e ............................................................ 87 Table 8. Cronbach’s Alpha for Measured Variable Sub-Indices and Factor I n d ic e s .............................................................................................. 90 Table 9. Factor Structure of the Cross-Sectional CFA Measurement M o d e l.............................................................................................................. 94 Table 10. Regression M odels (Low versus High Approach C o p in g )................... 96 Table 11. Time 1 Distress Outcome Regression Model with Interaction T e r m s ........................................................................................ 98 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Xlll Table 12. M ultiple Group M odel Development (Low versus High Approach Coping) ........................................................................................ 103 Table 13. Correlated Errors in the Final Single Group "Distress Outcome” M odels ........................................................................................ 105 Table 14. Summary of Final Two Group Structural Equation M o d e ls...................132 Table 15. Descriptive Statistics and Correlations of Measured Variables (SEM M odel Input) by Approach Coping G r o u p ................................................................................................................. 156 Ihble 16. Summary of Items Used in Subscale C o n stru ctio n ...................................157 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. XIV L i s t o f F i g u r e s Page Figure 1. Abstract M ediator and M oderator M o d e ls ................................................ 35 Figure 2. Structural and M easurement M o d e l............................................................ 68 Figure 3. M easurement Model in D e ta il ...................................................................... 73 Figure 4. Hypothesized Relationships Between Constructs ................................... 77 Figure 5. Distress Outcome M ultiple Group Structural M odel (Unstandardized) ............................................................................................158 Figure 6. Distress Outcome M ultiple Group Structural M odel (S tan d ard ized )..................................................................................................159 Figure 7. Fully Lagged M ultiple Group Structural Model (Unstandardized) ............................................................................................160 Figure 8. Fully Lagged M ultiple Group Structural Model (S tan d ard ized )..................................................................................................161 Figure 9. Change Scores M ultiple Group Structural M o d e l......................................162 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. XV Figure 10. 1990 Cross-Sectional M ultiple Group Structural M o d e l...................... 163 Figure 11. 1991 Cross-Sectional M ultiple Group Structural M o d e l......................164 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. XVI A b s t r a c t The present study hypothesized that the effect of coping on the relationship between life events (stressors) and psychological symptoms (distress) in adolescence depends on the type of coping skills used, specifically that avoidance coping skills mediate, and approach coping skills buffer or moderate, this relationship. As part of a large community-based prevention trial, a random sample of 9th, 10th and 11th grade students was administered a self-report survey that included life events, coping strategies, and psychological symptoms (anxiety, depression, and somatic complaints). Using baseline and one year follow-up data, longitudinal single and multiple group structural equation models were used to evaluate mediating and moderating effects of different coping strategies. Results supported the hypothesis that adolescent use of avoidance coping strategies of antisocial behavior, “partying” behavior, and substance use mediates the effect of stressors on increased psychological symptoms. The results also supported a stress-buffering effect of the use of high levels of approach coping strategies of exercise, information seeking, and relaxation. High levels of approach coping protect against the negative effects of stressful life events on psychological symptoms. Some gender and age-specific differences in levels of stressors and distress were observed. Females reported more stressful life events and more psychological symptoms or distress than males. Older students reported more Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. xvn stressful life events than younger students. The hypothesized results would suggest that prevention programs aimed at decreasing risk for mental illness, drug abuse, and problem behaviors in adolescents should emphasize both decreasing the use of avoidance coping strategies and increasing the use of approach coping strategies. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I n t r o d u c t i o n Twentieth century medicine and public health have conquered many of the traditional causes of morbidity and mortality, namely infectious diseases (with HIV being a notable exception). As this has occurred there has been an increasing focus on new concepts of health, disease prevention, and health promotion. Health is now viewed as m ore than merely the absence of disease and hence the biopsychosocial model of health promotion requires us to encourage the attainment of a state of complete physical, mental and social well-being. Furthermore, many states of ill health in any of the three domains can be traced to cognitive and behavioral patterns in individuals. Many of these patterns fonn during childhood and adolescence and maintain themselves later in life. M razek (1994) reviews a num ber of studies showing that although only a minority of adolescents experience m ajor depression (2-10%), several times these figures experience depressed mood during adolescence and that many cases of psychopathology identified as adult cases had their clinical or preclinical genesis during adolescence. With an improved understanding of how individual psychobehavioral differences influence psychological well-being researchers, healtii professionals, and educators may be able to better help individuals attain more satisfactory levels of social, psychological, and physical health. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 In particular, although stressful life events are known to increase the risk for depression, depressed mood, suicide, and suicidal ideation in both adolescents and adults (Kaplan & Sadock, 1991), there is a large degree of individual variation. This in turn leads to an interest in the person-level factors related to this difference in susceptibility. Furthermore, a prevention orientation emphasizes that early modification of behavioral and cognitive risk factors for depression is of greater value than later treatment of deep-seated psychological traits or characteristics (although research into some of these such as hardiness shows significant effects on an individual’s response to stress). This study examined the degree to which the person specific cognitive and behavioral factors of coping influence an adolescent’s experience of the psychological symptoms of anxiety, depression and somatization. In addition, it examines the extent to which specific coping factors mediate the relationship between stressful life events in adolescence and negative psychological symptomatology. Finally, because productive coping is expected to reduce the adverse psychological sequelae of stress, the study examined the extent to which specific coping methods moderate or attenuate the relationship between stressful life events and psychological symptoms. We examined these issues in a longitudinal sample of high school students as part of a larger, ongoing trial of the primary prevention of drug, alcohol, and tobacco use. As will be discussed later, this age group has specific developmental Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 3 characteristics that put them at specific risk for stressful life events and their negative consequences. B a c k g r o u n d a n d S i g n i f i c a n c e Health Behavior and Health Promotion Health Behavior Health behavior research can be divided into two domains, explanatory research and intervention research, with the first providing the substantive basis for the second (lessor, 1982a). lessor further states that the 'lack of theory- oriented research remains a major obstacle to progress in promoting adolescent health.” (lessor, 1982a) Theory provides the conceptual framework for understanding and making meaningful the particular health behavior of interest. In addition, it gives insight into whether specific behaviors are unique or rather are actually specific instances of a “theoretical category of mutually substitutable, alternative behaviors.” (lessor, 1982a p. 450) According to lessor’s Problem- Behavior Theory, there exists in adolescents a collection of behaviors including delinquency, early sexual intercourse, drug use, and school failure that are not independent behaviors, but rather manifestations of an underlying “problem- behavior” syndrome. lessor goes on to say (lessor, 1982a p. 450) that “[o]ne of the critical shortcomings of much contemporary health research is the slighting of attention to personality attributes or to relatively enduring individual differences. ” M ultiple studies have found that intra-personal level variables, that is, behavioral Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 4 and cognitive variables specific to an individual can account for over one third of the variance in a variety of diverse behaviors in adolescence including drug use, delinquency, and church attendance (multiple correlations over .60) (lessor & Jessor, 1977). Health Promotion According to Dryfoos (1990), the fundamental theoretical construct on which health promotion interventions are predicated is relatively simple. Interventions should be directed towards the common antecedents of the targeted problem or problems rather than at separate behavioral manifestations of the problem such as drug use, early sexual activity, and lack of academic achievement. Interventions can and should target both person-level and environmental antecedents. How then do we promote health, and what do we need to learn to do so effectively? According to Perry: [e]fforts to promote health can be divided into two main categories: those that are oriented towards weakening, reducing, and eliminating behaviors that compromise health; and those that are oriented towards introducing, strengthening, and reinforcing behaviors that enhance health. (Perry & Jessor, 1985 p. 134) W hat then, are some of the behaviors that compromise health in relation to stressful life events or enhance health in spite of stressful life events? And also, why focus on adolescents? W hat is unique about adolescence that makes high school students appropriate subjects for this study? Tkking the last first, once we have specific health promotion or disease prevention interventions in hand. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 5 adolescents are a relatively easy population to tai^et because school-based interventions will reach a very high proportion of them. Furthermore, changes in behavior and cognition caused by interventions will have a lifetime of cascading effects, in contrast to interventions with persons later in the life cycle. M ost important, however, are the particular developmental features of adolescence that make adolescent stress research so important. The other question above, what are the vulnerability or protective factors in the face of stressful life events, will be discussed in the sections on coping, below. Adolescent D evelopm ent an d Stress Several developmental theories bear on the question of why adolescence is a particularly stressful period during an individual’s life. O f the eight psychosocial developmental tasks which individuals experience during life (Erikson, 1968), identity formation, the fifth one, is the salient one during the 12 to 20 year old adolescent period. Identity formation is a complex task that requires sophisticated cognitive and social skills. Erikson says (1968 p. 87) that the task of identity formation “meets a crisis to be solved only in new identification with age mates and with leader figures outside the family.’’ During this search for identity, it is the exploration or the period of struggle and active questioning in making personal identity decisions in the domains of personal goals, values and beliefs which is potentially problematic for youths (Archer & Waterman, 1994; M arcia, Waterman, M atteson, Archer, & Orlofsky, 1993). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 Furthermore, individuals may progress through different domains of identity formation at different times: for example political ideologic choices may be made before religious or avocational ones. According to A rcher (1994), most true identity formation occurs in late adolescence (late teens) so those social resources and personal skills with which youths leave early adolescence and enter late adolescence are critical in successfully resolving identity crises during the path towards personal identity. In particular, adolescents must possess sophisticated decision-making, information gathering, and problem solving skills, as well as be able to evaluate alternatives and consequences. M any of these cognitive skills correspond to the fonnal or abstract phase in Piaget’s cognitive developmental theory (Piaget, 1972). During this phase adolescents develop the capacity for hypothetical-deductive reasoning, the ability to be introspective and think about their own thoughts, and they are able to understand the concept of probabilities. All of these skills are important for various aspects of the stress appraisal and coping process. Erikson (1968) said that identity fonnation is driven in part by needs for autonomy from parents and for increased affiliation with peer groups. Some more specific theories of adolescent behavior in general and in the domain of drug use and prevention would agree that an adolescent’s desire to emulate older peer behavior leads to vulnerability during his or her development. Specifically, lessor and others (lessor, 1982a; Pentz, 1985), as part of the larger Problem-Behavior Theory, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 7 have used the term “transition proneness” to specify the vulnerability to negative behavior that exists during the year of transition to a new school such as junior or high school. Specific stressors related to adolescent development include the issues of separation from parents, identification with peers, and individuation (Mechanic, 1983), and the issues of increased social and academic pressures (Jessor, 1982b; Johnson, 1986; Perry & lessor, 1985). These stressful and threatening situations and life events have been shown to be related to both physical and psychological well-being in both adults and adolescents (Billings & Moos, 1982; Blaney, 1985; Cohen, Tyrrell, & Smith, 1991; Compas, Wagner, Slavin, & Vannatta, 1986; Epstein & Perkins, 1988; Krantz, Contrada, Hill, & Friedler, 1988; Swearingen & Cohen, 1985). Preventive medicine and health behavior are concerned with understanding behavior and cognition related to health in order to design prevention programs to change health-related behavior. W hen considering preventive interventions directed at stress or its consequences it can be helpful to consider the concept of resilience. Resilience is the ability to recover from or successfully cope with significant stresses (Rutter, 1985). In particular, resilience involves an interaction of individual vulnerability, protective factors, and environmental stresses (Mrazek & Haggerty, 1994). .\s reviewed by Tkylor, some of the intrapsychic personality level vulnerability factors may include negative affective style. Type A behavior (particularly hostility) and others (Tkylor, 1990). O f more interest in terms of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. potential preventive interventions are the protective factors. Rutter (1985), writing from a psychiatric perspective, defines protective factors as “those factors that modify, ameliorate or alter a person’s response to some environmental hazard that predisposes to a maladaptive outcom e.” These protective factors fall into two categories: the positive effects of social support, and personality factors or personal characteristics that affect the individual’s ability to cope with stress (O’Grady & M etz, 1987). Individual level differences may be more amenable to modification through interventions, and coping skills rather than, for example, hardiness (Ouellette, 1993), are the individual level protective factor of interest in this dissertation. Although these stressful life events have been associated with increased psychological symptoms in adolescents, including somatic complaints, anxiety, and depression (Coddington, 1972; Johnson & McCutcheon, 1980; Mechanic, 1983; Swearingen & Cohen, 1985) significant individual differences exist. These differences exist in part because of differences in resources available and methods used to cope with stressful life events (Compas, 1987). Previous research with adolescents either: (1) used regression models which fail to control for measurement error when estimating causal relationships; (2) did not systematically compare whether coping has a moderator or mediator effect on the stressor-symptom relationship; or (3) did not consider whether these moderator and mediator effects are dependent on the type (approach or avoidance) of coping Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 9 strategy used (Holahan & Moos, 1991; M cCubbin, Needle, & Wilson, 1985; Pearlin & Schooler, 1978; Wills, 1985). This study attempts to address this concern by investigating the relationship of the person-level theoretical construct known as coping skills on the negative effects of stressful life events. Furthermore, because of the large amount of covariation in a variety of problem behaviors in adolescence (Bentler, 1987; Jessor, 1982a), new knowledge of person-level differences may ultimately be able to be leveraged into interventions in a variety of domains, either situation specific or situation independent. Some work of this type (Botvin, 1986; Botvin, Baker, Dusenbury, Tortu, & Botvin, 1990) suggests that life and social skills training (modifications of personal characteristics and skills) may be effective in the prevention of some problem behaviors. In order to understand these relationships, we must first understand what is meant by the terms stress, stressors, coping, and psychological distress. We must also examine the implications these concepts and relationships have for prevention and health behavior. Stress and Stressors Origins and Foundations In some sense the question "W hat is stress?” may seem ingenuous. After all, everyone knows what stress is— just turn on the television, read the newspaper, or browse the self-help section of a bookstore and it is obvious that Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 this is a word in common use. Yet research using term s with such common connotations is fraught with danger, for without further specification readers may believe they know what is meant by the term in question, yet the researcher may have quite different concepts in mind. For these reasons, some examination of the history of the term and its use in clinical and psychological research is in order. The English word “stress” has its origins in the M iddle English stresse and distresse in the 14th century. These in turn come from the Latin root distringere “to pull ap art.” (Mish, 1983) This fits with the common usage—stress tends “to pull apart” the stability of an individual’s environment. Physicists used the tenn in mechanics in the context of the load or force per unit area (stress) causing a deformation (strain) of the object under load (Hinkle, 1977). Clinically the concept, if not the precise term, has been a part of the physicians’ armamentarium for many centuries. Hans Selye, the father of m odem stress research, relates that Hippocrates, in the 4th Century B.C., describes disease as “not only suffering {pathos), but also toil {ponos), that is the fight of the body to restore itself toward norm al.” (Selye, 1976 p. 11) Later clinicians and physiologists expanded on this concept. For example, the idea that all living things maintain the constancy of their internal milieu (Bernard, 1858; Bernard, 1927) led to Cannon’s invention of the term “homeostasis” to describe the tendency of our bodies to be in a state of equilibrium between opposing forces (Cannon, 1915). Furthermore, Cannon also believed that external stress could Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 11 have an effect on the body through emotions (Cannon, 1915). In the 1950s, Wolff had several theses that were important to the buigeoning field. H e stated that (emphasis added): [T]he stress accruing from a situation is based in part on the way the affected subject perceives it; perception depends on a multiplicity of factors . . . . The common denominator of 'stress disorders’ is reaction to circum stances of threatening significance to th e organism . (Wolff, 1968 p. 8) Perhaps most famous of these conceptualizations of stress is Hans Selye’s General Adaptation Syndrome. This theory used the term stress in a specific, mostly physiologic sense. As with Wolff, this is a response-oriented model that defines stress as the “nonspecific response of the body to any dem and” (Selye, 1976 p. 55). Selye discusses at great length the difficulties he and others experienced with the terminology as he developed his theory and states that he later realized that he should have used the word strain since he was emphasizing the response, rather than the stimuli. Because of this initial error he was forced to invent a new word, “stressor” to describe the (antecedent) causes of the body’s responses (Selye, 1976 p. 51). The General Adaptation Syndrome has several stages: first comes an alarm reaction by the oiganism in response to the noxious stimuli; second, a stage of resistance during which the oiganism has adapted to the presence of the stimuli; and finally, a stage of exhaustion when the demands of the adaptation process are too great and maintained over too great a time (Selye, 1993). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 12 The fields of psychiatry and psychology also were interested in the stress concept. In fact, World War U provided a laige body of data representing a natural experiment into the reactions of people to stress. There was a profound interest in the individual differences in reactions to identical or sim ilar events. Sabshin (1972) discusses a 1945 book by Grinlcer and Spiegel, M en U nder Stress, that dealt with the responses of soldiers to combat, including those soldiers that responded with “shell-shock.” He states that because of this book the Department of Defense instated prim ary prevention interventions by psychiatrists and tried to improve their early detection of pathologic responses to stress (secondary prevention) (Sabshin, 1972). Studies of the experiences of concentration camp survivors (Eitinger, 1973; Eitinger & Major, 1993) and starving ghetto residents in Warsaw (Ibshnet, 1966) also emphasized individual differences in responses to stress. For example, the physicians in the Warsaw ghetto, starving themselves, gave food to their patients and managed to conduct research on starvation—research that was proscribed by the Nazis (Tlishnet, 1966). Psychosomatic medicine of the 1940s and 1950s emphasized the concept of illness specific personalities, i.e. the “ulcer personality” (Alexander, 1950). W hen this concept was later disproved, psychosomatic medicine itself lost favor. It began to enjoy a resuigence as the emphasis on purely intrapsychic factors was replaced with a recognition that environmental factors could cause psychological consequences which in turn could cause illness (Lazarus & Folkman, 1984). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 13 Entire fields developed, including those of psychoneuroimmunology (Kiecolt-Glaser & Glaser, 1992; O’Dowd, 1988; Ratliff-Crain, Temoshok, Kiecolt-Glaser, & Thmarkin, 1989) and health psychology (Stone, Cohen, & Ader, 1979). Stressor Theories One stress theory categorized stressors by magnitude and extent. That is, m ajor changes that affect laige numbers of persons, m ajor changes that affect one or a few people, and daily hassles (Lazarus & Folkman, 1984). The first type would be the type studied by Thshnet and Eitinger (Eitinger, 1973; Eitinger & Major, 1993; Thshnet, 1966) and would include natural disasters and wars. The second type would include the death of a loved one, an event beyond one’s control, and events such as major examinations, over which one has a certain amount of control. The final type would include breaking the coffee pot in the morning, fighting traffic on the way to work, or aiguing with a coworker. Another categorization referred to by Lazarus is the Institute of M edicine’s duration-based categories (Elliott & Eisdorfer, 1982). These include acute stressors such as being in a near accident in your automobile, stressor sequences deriving from a single antecedent event such as a divorce, chronic intennittent stressors such as tax season, and chronic stressors such as a difficult marriage or a high-pressure job. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 14 The field of stress research, particularly the quantitative aspects, was galvanized by Holmes and Rahe’s development of a theory of stress based on the concept of readjustment (1967). Their interest was in clinical illness and in the stressful life events that were associated with illness. By studying over 5000 patients and eliciting from them reports of stressful life events that occurred prior to the onset of their illness, they developed the theory that the true indicator of a stressor was that it required readjustment on the part of the individual who experienced it. Important components of their theory included that a predetermined set of stressful life events could capture much or most of the stress that diverse individuals experienced. Second, that the amount of readjustment necessary for a given event did not differ significantly between individuals. And third, that even events that could be considered positive or happy, such as a holiday or vacation, should be considered stressful because they require readjustment or adaptation on the part of the individual. A major criticism of the approach of Holmes and Rahe (or their followers) has been that no distinction has been made between events of putatively positive or negative valence (Johnson & M cCutcheon, 1980; Sarason, Johnson, & Siegel, 1978). Holmes and Rahe seem to believe, as did Richard Hooker, a 16th century English theologian, that “change is not made without inconvenience, even from worse to better." Antonovsky (1979) would say that some of the events measured by Holmes and Rahe’s scale are actually health promoting, salutogenic factors Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 15 rather than stressors. However, because of our concern with intrapersonal factors related to the stress response we believe that “positive” events requiring adjustment are not qualitatively different from negative life events and hence should not have a different score or dimension attached to them. Furthermore, Wills (1985) among others failed to show support for a positive effect of positive events. Another category of stressor theories proposes that although the readjustment theory is essentially correct, stressors are more properly categorized in several distinct domains. Several groups have found relatively consistent sets of factors of stressful life events (Rahe, Pugh, Erickson, Gunderson, & Rubin, 1971; Skinner & Lei, 1980). A laige group of researchers support and test what is known as the “Daily Hassles and U plifts” theory of stressful life events (Kanner, Coyne, Schaefer, & Lazarus, 1981). This theory holds that the day to day annoyances of life are have more psychological impact on an individual than the major life events of the SRE. These events include a variety of internal as well as external types of events and consist of the “ irritating, frustrating, distressing demands and troubled relationships that plague us day in and day out” (Lazarus & Delongis, 1983). Although Lazanis and Delongis and others in their group feel that this is a more valuable theory of stressful life events than Holmes and Rahe’s, as will be discussed below, serious criticisms have been raised of it—specifically that it is Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 16 confounded with many of the most common outcome measures, including psychological distress (Dohrenwend & Dohrenwend, 1974). Stressor Measures The readjustment based theory of Holmes and Rahe (1967), and the study that spawned it, led to an instrument, the Social Readjustment Rating Scale (SRRS) consisting of a list of 43 stressful life events. The significant elements of their theory had implications for their measure in that it was a closed-end instrument, both positive and negative events were included (e.g. both marriage and death of a spouse), and each event was given a rating by independent experts of the severity or stressfulness of the event. This process is reviewed by Rahe (1974) and discusses the associations he has found between the SRRS and its successor the Schedule of Recent Events (SRE) and clinical illness in both retrospective and prospective studies. In addition, some researchers have found support for a domain-specific formulation of stressful life events, e.g. separate clusters of job, relationship, and personal stressful life events (Newcomb, Huba, & Bentler, 1981; Rahe, 1974; Skinner & Lei, 1980). There are problems with the methodology used to derive the original life events used in Holmes and Rahe’s scale and with the retrospective studies in particular. These criticisms include most notably the possibility of a memory bias on the part of those with illness. The scale has the advantage, however, that it measures external events not easily confounded with intrapsychic phenomena (in Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 17 contrast to some of the other stressor measures). Furthermore, the current body of evidence supporting the relationship between stressful life events and psychopathology and pathophysiology is so great that it is clear that the instrument is a valuable one (Antoni, LaPerriere, Schneiderman, & Fletcher, 1991; Antoni, Schneiderman, Fletcher, Goldstein, Ironson, & Laperriere, 1990; Dohrenwend & Dohrenwend, 1974; Irwin, Daniels, Bloom, Smith, & Weiner, 1987; Kellner, 1991). Many others have followed the Holmes and Rahe for at least a portion of their stress assessment. For example, one part of M oos’ Health and Daily Living instrument is a schedule of recent events (Moos, Cronkite, Billings, & Finney, 1984). Sometimes the SRRS (or SRE) is modified for specific populations, such as the children of alcoholics (Roosa, Sandler, Gehring, Beals, & Cappo, 1988) or for children of various ages (Coddington, 1972) or for adolescents (Newcomb et al., 1981). Coddington’s work is of particular interest in this dissertation because he developed different lists of life change experiences for each of several age groups: preschool, elementary school, junior high school, and senior high school. To the extent that these instruments do not stray from the theory of stressful m ajor life events of an external nature, they are quite useful for specific populations. As discussed above, the modification of readjustment theory that postulates that stressors assort themselves into distinct areas is termed the domain specific Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 18 readjustment theory. Instruments based on this theory include lists of life events similar to those in the SRE, but then reported as separate scores in separate domains. Skinner (1980) also found that the S SE items could be separated into six relatively homogeneous domains: personal and social activities, work changes, marital problems, residence changes, family issues, and school changes. Finally, of most import for this dissertation, Newcomb, Huba, and Bentler found that multidimensionality existed in the life events experienced by adolescents as well, although their labels and even their specific events differed: family/parents, accident/illness, sexuality, autonomy, deviance, relocation, and distress (Newcomb et al., 1981). Some stress measures are modifications of the SRE made for substantive or measurement related reasons, rather than to individualize the instrument for a specific population. These include Sarason, Johnson, and Siegel’s (1978) Life Experiences Survey and Johnson & M cCutcheon’s Life Events Checklist (1980) use positive and negative valences when scoring to address the criticism that directionality of change is important. Other suggestions for modifications of the SRE have included "standardizing” the score by dividing an individual’s score by the number of listed events that are applicable to the individual, e.g. if an individual is not married, then the “divorce” event does not apply (Ander, Lindstrom, & Tibblin, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 19 1974). A nother modification of the SRE is to add a weighting for the recency of the event (Horowitz, Schaefer, & Cooney, 1974). The Daily Hassles and Uplifts theory measures hassles with a portion of the total scale consists of 117 hassles including misplacing and losing things, concerns about money, not enough tim e for family, and others (Kanner et al., 1981). Each item also had a severity rating given, from 1 to 3. The scale yielded three scores: a sum of the num ber of hassles experienced, a sum weighted by severity, and an average severity or intensity score. They also proposed that daily uplifts, or events with a positive psychological valence, would be have a beneficial effect on psychological outcome measures. T heir work did support the utility of this conceptualization of stressors rather than that of m ajor life events. That is, hassles were more strongly related to psychological symptoms than were life events, after adjusting for the effect of hassles, the association of life events and symptoms was not significant, and that after adjusting for the effect of life events the association of hassles and psychological symptoms remained significant. However, using clinical psychologists as raters, Dohrenwend, Dohrenwend, Dodson, and Shrout (1974) found that 37 of the 117 Hassles items were seen as more likely than not to be signs of psychological symptoms and an additional 53 were at least somewhat likely to be symptoms. This confounding of predictor and outcome measures makes the results reported by Kanner, Coyne, Schaefer, and Lazarus (1981) difficult to interpret. Brantley, Waggoner, Jones, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 0 and Rappaport’s 1987 development of the Daily Stress Inventory supports the daily hassles theory, but adds an individual severity rating to each item. Finally, we have Cohen’s group, who developed a Perceived Stress Scale that includes items that seem to tap dimensions including hassles, self-efficacy, feelings of control, and perhaps psychological symptoms such as depression or anxiety (Cohen, Kamarck, & M ermelstein, 1983). Stressful Life Events and Adverse Outcomes The num ber of books, articles, and review articles describing some aspect of the stress process is immense. In 1976 Selye wrote that in the preceding forty years since he had first introduced the General Adaptation Syndrome and the concept of diseases of adaptation more than 110,000 articles and books had been published on the subject (Selye, 1976 p. xi) with 1,500 articles and thirty books published by Selye himself. Adverse outcomes studied in relation to stressful life events include clinical outcomes and pathopsychologic outcomes. The latter includes both clinically diagnosable psychiatric disorders as well as a broader constellation of psychological symptoms sometimes termed psychological distress. Stress has been associated with immune function in numerous studies (Antoni et al., 1991; Glaser & Kiecolt-Glaser, 1988; Gorman & Kertzner, 1990; Irwin et a l., 1987; Irwin, Daniels, Bloom, & Weiner, 1986; Kessler, Foster, Joseph, Ostrow, Wortman, Phair et al., 1991; Kiecolt-Glaser, Glaser, Shuttleworth, Dyer, Ogrocki, & Speicher, 1987; Schlesinger & Yodfat, 1991; Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 21 Workman & La Via, 1991). Although usually these immunologic studies have not established a direct link to disease, their insights into possible biologic mechanisms whereby stress might exert a disease producing or promoting effect are valuable. For additional insight into the neurochemical mechanisms of the stress reactions see the recent review by McEwen (1993). It has been associated with the onset of the common cold (Cohen et al., 1991) in a study that prospectively infected individuals with different levels of self-reported stress with one of several types of viruses that cause the common cold. A laige body of work relates coronary heart disease to stress, two recent reviews in psychological journal include those by Epstein and Perkins (1988), and by Krantz, Contrada, Hill and Friedler (1988). In the psychological domain, there are also numerous studies associating stressors, including stressful life events, with psychological outcomes. These studies include both longitudinal and retrospective studies and are published in both psychological and psychiatric Journals. Blaney (1985) provides one review of the stress and depression literature in adults, noting both experimental research, where the stressor is often artificial; natural experimental research, where the focus is on the effect of a specific major stressor such as the death of a loved one; and survey research, where the stressors are life events—whether measured retrospectively or prospectively. An example of longitudinal survey research in adults is the work of Billings and Moos (1982). They note specifically the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 2 potential problem of biased life event reports from respondents experiencing significant psychological symptomatology and recommend using multivariate techniques (which they do) and controlling for measurement error (which they do not). Among adolescents, several researchers have found in both cross-sectional (Brown, 1989; Hawkins, Hawkins, & Seeley, 1992) and longitudinal studies (Compas et al., 1986; Swearingen & Cohen, 1985) that stressful life events are associated with increased psychological symptomatology (Coddington, 1972; Johnson & McCutcheon, 1980; Mechanic, 1983). Coping Origins and Foundations The English word coping is derived from the Old French coup “blow,” from the Latin colaphus and Greek kolaphiis “to buffet”—that is, its roots are in defending in battle (Mish, 1983). The same dictionary defines coping as “dealing with and attempting to overcome problems and difficulties.” (Mish, 1983) Coping research has two broad theoretical and research foundations. The first foundation corresponds in part to the etymology given above and rests on Selye’s theories and Darwinian thought, in which coping is a survival mechanism that enables an oiganism to avoid, control, or overcome noxious environmental agents. This model of coping fails to take into account the cognitive and emotional richness that is characteristic of human coping and so is unduly simplistic (Lazarus & Folkman, 1984). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 23 The second foundation derives from the psychoanalytic concepts of ego- defenses against threats to psychological integrity. In the traditional psychoanalytic formulation, coping refers to the most advanced or mature ego- defense mechanism, with neurotic and psychotic modes of ego-defense representing less mature ego processes (Kaplan & Sadock, 1991). Associated with each of the immature ego-defense mechanisms was a specific psychopathology, e.g. hysterical neuroses were linked to repression (Lazarus, 1993). This hierarchical model of coping emphasizes coping as a style or trait and has produced a number of trait-like measures of coping contrasting aspects of the hierarchy. Cohen (1987) lists several measures of this type contrasting repression, avoidance, or denial with sensitization, isolation, or intellectualization. Coping Theories Clinically, however, the links between hierarchically expressed styles of coping and psychopathology were not as tight as theory might iiave predicted and therefore, just as stress researchers moved into other areas when the theories of 'ulcer personality” did not bear up under scrutiny, coping researchers also broadened and modified their models. The outstanding characteristic of these new models was a view of coping as a process. Several researchers in the late 1970s developed models and measurements of coping as a process which changes over time (Billings & Moos, 1982; Folkman & Lazarus, 1980; Pearlin & Schooler, 1978; Stone & Neale, 1984). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 24 Lazarus and Folkman (1984) define coping as the process of managing external or internal demands that are perceived as taxing o r exceeding a person’s resources. This assessment or appraisal stage, is important to the coping process for coping only occurs after a primary appraisal of a situation as hannful, threatening, or challenging. Secondary appraisal occurs when an individual evaluates his or her external resources such as money or time, internal resources such as coping skills, and then determines whether they will be sufficient to overcome the threat posed by the particular situation (Lazarus & Folkman, 1984). Notice also that a distinction is made between coping, which involves purposeful or effortful responses and reflex or instinctive responses to stress (Compas, 1987; Lazarus & Folkman, 1984). The appraisal process is particularly critical in adolescence, for it is during this time that the cognitive skills necessary for primary and secondary appraisal are being refined and developed (Erikson, 1963; Erikson, 1968; Piaget, 1972). Function of Coping Folkman and Lazarus’s theory distinguishes between problem-focused and emotion-focused coping responses (Folkman & Lazarus, 1980; Lazarus & Folkman, 1984). In general they have found that both sets of strategies are used in most stressful situations (Folkman & Lazarus, 1980), but problem-solving strategies may be more successful for managing controllable stressors and emotion-focused coping may be more appropriate for uncontrollable stressors. In Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 25 refining the theory, Lazarus identifies eight specific coping factors: confrontive coping, distancing, self-controlling, seeking social support, accepting responsibility, escape-avoidance, planful problem solving, and positive reappraisal (Lazarus, 1993). Others have identified similar, though not identical factors. M odes or M ethods of Coping Another theory of coping, while similar to the emotion-focused/problem solving oiganization, relies on the contrast between active (or approach) coping and avoidance coping (Billings & Moos, 1982; Holahan & Moos, 1991; Moos, Brennan, Fondacaro, & Moos, 1990). The distinction made is based on the focus of coping: (1) active coping that involves both behavioral and cognitive efforts to directly address the stressful problem versus (2) avoidance oriented coping that attempts to reduce tension by escapist behaviors and thoughts. M ore recently. M oos and his group have divided each of these two methods of coping into cognitive and behavioral axes (Moos & Schaefer, 1993). Yet another theoretical model, that of Pearlin and Schooler (1978), divides coping strategies into three groups. Some coping responses seek to modify the situation that represents the source of stress—these would in general be classified as active, behavioral coping strategies. A second type of coping seeks to minimize the significance of the stressful situation. This would include reassurance or telling oneself the problem isn’t that important. The third type of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 2 6 coping strategy tries to minimize the negative response to the stress, e.g. trying not to worry, relaxing, accepting the problem. Finally, Compas’s theoretical formulation distinguishes between coping resources, styles, and specific coping efforts (Compas, 1987). Resources would include some elements included above under social support (i.e. the availability of a supportive social network), others under personality characteristics (e.g. positive self-esteem), and other more specific coping skills such as problem solving ability. Coping styles refer to methods of coping that are common across problem domains and across time. Some of this commonality may reflect underlying beliefs or values or a particular personality trait. However, having a specific coping style does not necessarily imply the presence of a particular personality trait (Compas, 1987). Lastly, specific coping efforts are the cognitive or behavioral actions taken in response to a specific stressful life event. Because specific situations likely require different coping responses, some of these theories seem less useful than others from a prevention perspective. In particular, approach/avoidance theory seems more useful than problem/emotion- focused theories because productive coping responses could easily come from either problem-focused or emotion-focused domains. This threatens the construct validity of the problem- or emotion-focused coping constructs, because both of these types of coping will have a productive and a nonproductive component. Covariation of these constructs with positive and negative outcomes will hence be Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 27 attenuated, making them less useful as predictors of outcomes such as psychological distress, drug use, or other problem behaviors. In contrast, Billings and M oos have found that productive coping is more likely to use the approach rather than avoidance strategies (Billings & Moos, 1981). Coping Measures Cohen gives an excellent categorization of coping dimensions assessed by some commonly used measures (1987 Tkble 1, pp. 288-289). The major divisions include trait measures, consisting of the unidimensional schema of Repression- Sensitization and Coping-Avoidance and multidimensional schemes measuring the more classic psychoanalytic concepts of defense mechanisms or coping/defense mechanisms. The second m ajor group she terms episodic measures, including unidimensional measures such as the Avoidance-vigilance interview and the Denial Scale. M ore interesting in the context of this work, considering the complexity of the underlying behaviors and cognitions, are the multidimensional measures. These can measure functions of coping, modes of coping, or a more diverse set of coping factors. Folkman and Lazarus (1980) developed the revised Ways of Coping Questionnaire (WOC). After identifying a specific stressful situation, for each of sixty-six specific coping responses individuals are asked to rate on a . four point Likert-type scale their use of each coping response. This scale also has eight subscales: positive reappraisal and accepting responsibility (cognitive approach). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 28 seeking support and problem solving (behavioral approach), distancing (cognitive avoidance), and escape-avoidance (behavioral avoidance). The eight factors are listed according to how M oos (1993) would categorize them , Lazarus divides them into emotion-focused (distancing, escape-avoidance, positive reappraisal, and accepting responsibility) and problem-focused (seeking support, problem solving). Pearlin and Schooler (1978), in one of the earliest coping instruments developed, interviewed 100 adults in Chicago to generate coping strategies that they then used in closed-ended instrument. As mentioned above, one of their m ajor contributions is to categorize coping strategies based on whether the strategy tries to modify the problem (similar to problem-focused coping) or the significance of the problem to the individual (sim ilar to emotion-focused coping). They further divided coping into 17 factors that depended on the problem area in addition to the dimensions above. Other measures of coping include interview instruments (Zautra & Wrabetz, 1991), and open-ended questionnaires. Stone and Neale (1984), in developing a coping instrument for repeated daily use, divided their 55 item measure into eight categories: distraction, situation redefinition, direct action, catharsis, acceptance, social support, relaxation, and religion. Spirito (1988) developed a brief instrument, Kidcope, for use with pediatric populations. Subjects are asked to select a problem they encountered in the prior month and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 29 rate each of 10 coping strategies on the frequency with which they used it and how effective they thought it was for them. They found moderate test-retest correlation coefficients and demonstrated reasonable concurrent validity by showing moderate to high correlations with other coping scales. M easuring coping along the avoidance/approach axis, the adult form of the Coping Responses Inventory (CRI) consists of 48 items that measure eight coping factors: logical analysis and positive reappraisal (cognitive approach coping), seeking guidance/support and problem solving (behavioral approach coping), cognitive avoidance and resigned acceptance (cognitive avoidance coping), and seeking alternative rewards and emotional discharge (behavioral avoidance coping). Moos and his group have also developed specialized versions of this instrument for use with adolescents (Ebata & Moos, 1991; Moos, 1990) and with coping by health care workers (Schaefer & Moos, 1991). Overall, the Ways of Coping scale and the Coping Responses Inventory included in the M oos’s (1984) Health and Daily Living inventory are the most widely used instruments and have the most substantive and theoretical underpinnings. One issue that arises has to do with how the coping items are worded. Some instruments ask the respondent to think of a specific problem and then to tell how much each coping method was used for the particular problem. This has the advantage of making a concrete example of coping salient to the respondent and may serve to prompt or jog an individual to remember coping Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 30 responses which would otherwise be forgotten. However, instruments that instead ask for “usual methods of coping” may eliminate the possibility of distortion because of very situation specific recall. Cooing and Rvchologicat Outcomes Some of the seminal work on coping focuses only on coping as a process related to stressful life events, particularly the more formative theoretical work. However, many other studies use specific outcome variables as the endpoint of interest. Outcomes examined are as diverse as adolescent sports injuries (Smith, Smoll, & Ptacek, 1990) and HIV infection (Blaney, Goodkin, M oigan, Feaster, M illon, Szapocznik et al., 1991). However, much of the work uses psychological outcomes as the endpoint of interest. Some researchers categorize their research in classical stress (life events) and strain (psychological symptoms or outcomes) terms (Pearlin & Schooler, 1978). Rutter (1985) provides a psychiatric perspective on the relationship between coping and other resiliency factors and psychiatric disorder. Several researchers have examined the relationship between coping and psychological symptoms or distress in adults, with Zautra and Wrabetz (1991) and Holahan and M oos (1991) being two representative, longitudinal studies. In both of these studies coping was modeled as a direct antecedent of psychological distress—that is, a direct effect between coping and psychological distress was tested for (and found). A simpler study design was used by Coyne, Aldwin, and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 31 Lazarus to show that depressed persons differ from non-depressed persons in the types of coping strategies they use (Coyne, Aldwin, & Lazarus, 1981). Folkman and Lazarus (1986) also found differences on some coping measures between depressed and non-depressed individuals. Compas, both in his own work (Compas, M alcame, & Fondacaro, 1988) and in a review (Compas, 1987) discusses how children and adolescent’s cope with stress and coping’s role in reducing adverse psychological states associated with stress. One of his findings (Compas et al., 1988) was that children and adolescents, like adults, tend to use multiple coping strategies when dealing with stressful life events. Some of his results relating specific types of coping with emotional and behavioral problems were contradictory and unexpected—for instance that coping with social problems seemed related to emotional and behavioral problems but coping with academic problems was not. He concluded that further work with these populations is warranted. A possible explanation for these apparently contradictory results may relate to the inability of the models tested to distinguish between mediating and moderating models. Psychological Sym ptom s an d Distress Several methods exist for the identification of individuals with various levels of psychopatliological symptoms. The primary method is the diagnostic interview. In children and adolescents the traditional view held that an unstructured interview was best because the cognitive and behavioral demands of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 32 a structured interview demanded too much of the youths (M okros & Poznanski, 1992). However, when the goal is to identify symptomatology and not to make definitive diagnoses, structured self-report instruments have several advantages. In clinical practice, easily administered self-report questionnaires are a low-cost case-finding method that can allow physicians to better detect patients with depressed mood or other affective disorders (Depression Guideline Panel, 1993). Although they often have high false positive rates when used for clinical diagnosis (e.g. positive predictive values in the neighborhood of 20% for m ajor depression in contrast to the point prevalence of approximately 7%) they can be a useful screening tool (Depression Guideline Panel, 1993 p. 76). In the context of health behavior research, particularly population-based research with laige sample sizes, these instruments allow research into psychopathology to proceed without prohibitive expense. In addition, Derogatis enumerates a num ber of characteristics of self- report scales in the context of stress research (Derogatis, 1982; Derogatis, Lipman, Rickels, Uhlenhuth, & Covi, 1974). The economy mentioned above comes because of several characteristics of self-reports. They are usually brief, do not require a professional to administer, and are easy to score. Another advantage, often not thought of, is that the self-report elicits from the individual himself information about his experience of a particular phenomena, in this case depressed or anxious mood. Clearly individuals may distort their reports for a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 33 variety of reasons, but many of the purported problems in this regard do not seem to be significant in actual administration of these scales to a variety of populations (Derogatis, 1982). The num ber of scales which measure psychological symptoms is very large, and only a few of the most common will be mentioned here. Some of the unidimensional psychological symptom inventories include the Beck Depression Inventory (Beck, Ward, Mendelson, M ock, & Erbaugh, 1961), the Zung Self- Rating Depression Scale (Coulehan, Schulbeig, & Block, 1989), the State-Trait Anxiety Inventory (STAI) (Spielbeiger, Gorsuch, & Lushene, 1970), and the C enter for Epidemiologic Studies Depression Scale (CES-D) (Radloff, 1977). An example of a multidimensional scale is the Hopkins Symptom Checklist (HSCL) (Derogatis et al., 1974) and its successor the SCL-90 (Derogatis, 1975). Although each of the unidimensional scales has aspects to recommend it, the SCL-90-R has the advantage that it assesses multiple dimensions of negative affect states and psychological symptomatology in a relatively brief format. Furthennore, it has become enough of a standard in multidimensional assessment of psychological distress that in recent years several studies have used it as the gold standard for construct validation of newer, more specific scales (Derogatis & Coons, 1993). Derogatis and his colleagues have shown that the SCL-90-R has high levels of test-retest and internal consistency reliability. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 34 In psychiatric populations the SCL-90-R has been shown to accurately detect depression and to be able to identify differential levels of symptomatology. In addition, Derogatis lists over 500 studies have used the SCL-90-R (Derogatis, 1990). The SCL-90-R has been used in a num ber of studies of stress (Dohrenwend, Dohrenwend, Dodson, & Shrout, 1984; Horowitz, Kmpnick, Kaltreider, Wilner, Leong, & Marmer, 1981). The precursor to the SCL-90-R, the HSCL, assessed five symptom dimensions: somatization, obsessive-compulsive, interpersonal sensitivity, anxiety and depression (Derogatis et al., 1974). The SCL-90-R, in addition to slightly modifying some of the existing 58 items, adds items to measure the symptoms/dimensions of hostility, phobic anxiety, paranoid ideation, and psychoticism. However, depression and anxiety are the negative mood states most identified with reactions to stress and their measurement is emphasized in this study. Mediator versus Moderator Distinction Results of some of the studies of the relationship of coping to adolescents’ response to stress have been inconsistent. In addition, a number of intervention studies have used specific coping strategies such as drug use or aggression as outcome variables without assessing the simultaneous impact of other coping strategies. However, adolescents often use multiple coping strategies as a response to stressful life events. Evaluation of the mediating and moderating Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 35 effects of specific coping strategies simultaneously would clarify the role of coping in adolescent stress. This study makes explicit the distinction between coping’s mediation and moderation of the relationship between stressors and psychological distress. In this study we hypothesized that certain coping strategies, tenned avoidance coping, mediate the effect of stressors on psychological symptoms. Other coping strategies, collectively termed approach coping, were hypothesized to act as moderators of the relationship between stressors and psychological symptoms. I n g e n e r a l . M ediator vs M oderator M odel SD SC Distres ■*1 Coping, M ediator M odel Stressoi M oderator M odel Distn mediator and moderator relationships can be expressed as the sets of relationships shown in Figure 1 on page 35. A mediator variable accounts for the relation between the predictor F igure 1. Abstract M ediator and M oderator M odels variable, in this case stressors, and the outcome variable, in this case psychological symptoms. If the research design supports strong causal inferences, mediator variables explain the mechanism by which the predictor variable exerts its effects on the outcome Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 36 variable. In the top diagram in Figure 1, a simplified mediator model is presented. This model is termed a completely mediated model if the parameter labelled SD is nonsignificant because all of the effect of the predictor variable (stressors) on the final outcome variable (distress) occurs through changes in the mediator variable (coping). An additional model, termed partly mediated would contains a significant direct path, SD, from stressors to distress. We designate the paths with letters, SC for the stressor-coping path, CD for the coping-distress path, and SD for the stressor-distress path. For coping to be a mediator, (1) variations in levels of stressor must significantly account for variations in the mediator (path SC), (2) variations in the mediator must significantly account for variations in the dependent variable (path CD) (Baron & Kenny, 1986), and (c) the mediated effect, or the product of SC and CD, must be significant. (1) / CD~Ss c +SC~Sc d +SscScd Prior to 1982, appropriate statistics to test for mediation effects were unavailable (MacKinnon, Johnson, Pentz, Dwyer, Hansen, Flay et al., 1991). Judd and Kenny (1981) suggested comparing program effect estimates with and without the inclusion of the mediating variable in the regression models. If the program effect estimate is substantially reduced by the inclusion of the mediating variable in the model, then there is evidence that the program effect is mediated by the variable. In 1982, however, Sobel (1982; 1986) derived the asymptotic Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 37 standard error of this mediated effect and so we can use the z-statistic given in equation (1) to test for a significant mediation effect. We test for a significant mediator by testing the significance of the mediated or indirect link between stressors and psychological symptoms through (mediated by) avoidance coping (Dwyer, M acKinnon, Pentz, Flay, Hansen, Johnson et al., 1989; MacKinnon et al., 1991). In equation (1) small s refers to the standard error of the parameters given by the subscripts, and SC and CD refer to the param eter estimates of the given paths. The num erator is the mediated effect and the denominator is Sobel’s (1982; 1986) derivation of the standard error of this mediated effect. Therefore, the mediated effect divided by its standard error yields a z-statistic with standard properties. Two prim ary methods have been used to evaluate mediator and moderator models, regression and structural equation modeling. Using regression, we would regress the mediator on the predictor variable, and the outcome variable on both the mediator variable and the predictor variable (Dwyer et al., 1989; MacKinnon et al., 1991). With structural equation modeling (SEM), all of the parameters of interest can be estimated at once. In addition, because measurement error will attenuate the relationships found, SEM has the potential to provide more accurate estimates of these relationships by correcting for measurement error (Newcomb & Harlow, 1986). It should be noted that measurement error in the mediator Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 38 variable will tend to cause underestimates of the mediated effect and hence overestimates of the direct effect (stressor-distress) (Baron & Kenny, 1986). Distress=a +p ^Stressors*^ ^Coping+p ^{Stressors*Coping) (2) The classic fonnulation of a moderator variable is a variable that affects the direction and/or strength of the relationship between an independent or predictor variable and a dependent or outcome variable (Baron & Kenny, 1986). For example, in the context of this study, on hypothesis was that for individuals who do not use approach coping skills, the relationship between stressors and psychological distress will be stronger than for those who do use approach coping skills. Tests for moderator effects must both identify a difference in relationship between the independent and dependent variable as a function of the moderator and test this difference for significance. Common tests for moderator variables assess the significance of a multiplicative term in a regression equation (Blaney et al., 1991; Cohen & Wills, 1985 for a review; Cronkite & Moos, 1984; Kliewer & Sandler, 1992; Wills, 1986; Windle, 1992). For example, if we were regressing distress on stressors and looking for a moderator effect we would use the regression equation given in equation (2). M ore specifically, the regression coefficient 83 is the coefficient of interest for the moderator effect—if it is statistically significant then we can conclude that a significant moderator effect exists. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 39 Methods of Testing As noted above, moderator and mediator models are typically tested using two distinct methodologies. The use of regression was reviewed above. An alternative method is to use structural equation modeling. Structural equation modeling explicitly models the measurement error inherent in most measures and particularly in measures of complex behaviors and cognitions such as coping and psychological symptoms. Constructs of interest are measured by groups of observed variables, whose covariation is modeled as occurring because of their relationship to the underlying construct. Covariation between variables which measure different constructs are modeled as occurring because of covariation between the constructs. This allows the separation of issues of measurement for example, issues of internal consistency and test-retest reliability from those of conceptual stmcture, for example, measurement construct validity (Cook & Campbell, 1979). A detailed explanation of the model creation, fitting, modification, and interpretation process is describe in the Methods section. Structural equation modeling allows us to simultaneously test for the moderator effects listed above, as well as test for differences of other types based on levels of the moderator variable (Holahan & Moos, 1991). For example, we tested for the measurement invariance of the latent variables between groups differing on levels of the moderator variable. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 40 Conceptual Models Several conceptual models can be evaluated using SEM to assess the interrelationships between stressors, coping, and psychological symptoms. M oderator models are tested in SEM using a multiple group approach, which means that models are fit simultaneously using groups defined by levels of the moderator variable. Significant parameter differences between the groups suggest a m oderator effect. All of the conceptual SEM models tested in this study are two group models, based on high versus low levels of Approach Coping. Cross-Sectional Models In the simplest model, similar to the mediated model in Figure 1 on page 35, stressful life events predict avoidance coping, which predict psychological distress. A direct effect from stressors to distress is also hypothesized and estimated in the model. Because the time lag between use of specific coping strategies and effects on psychological distress is not known, it is not clear what measurement lag is most appropriate. If the actual time for coping effects to be maximal is much shorter than one year, for example one month, then it may be that coping measured at the same time as distress has less measurement error compared to the true measure (coping one month ago) than would a coping measure from one year ago. This provides some justification for assuming a regression effect between Avoidance Coping and Distress measured at the same time. The cross-sectional Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 41 models assume regression effects exist between Life Events and Avoidance Coping as antecedents and Distress as consequent, all measured at a single time. A regression effect between Life Events and Avoidance Coping is also modeled. As discussed previously, because of the wording of the Life Events items, they clearly represent events occurring prior to other variables measured in the same questionnaire. Distress Outcome Longitudinal M odel If the effects lag between coping and distress is closer to one year, then longitudinal models are more appropriate for testing. The simplest longitudinal transformation of the cross-sectional models, using only two waves of data, would use Life Events and Avoidance Coping at Tune 0 to predict Distress at Tune 1. To more accurately assess the effect of Life Events, Life Events at Tune 1 was added to the model (this construct represents stressful life events in the year between measurement points as opposed to T m e 0 Life Events that represents events occurring in the year prior to baseline measurement). To further strengthen the model we controlled for baseline Distress levels. The Distress Outcome M odel (shown in unstandardized form in Figure 5 on page 158 and in standardized form in Figure 6 on page 159) is a structural or path model including regression effects representing unidirectional influences of one factor upon another across time. In general, we did not include regression paths within a single time because of the difficulty of making causal attributions from cross Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 42 sectional data. Withir.-time associations were modeled as correlations between constructs or between residual variances. However, because the life events item explicitly asks about stressful life events from the previous twelve months, the life events construct from 1990-91 (Time 0) is presumed to antedate all other constructs including other 1990-91 constructs and the 1991-92 (Time 1) life events construct is presumed to antedate the other 1991-92 constructs. M ore details are given in the Results section. Fuliv Lagged Longitudinal M odel The most complete conceptual model is termed the Fully Lagged Model (shown in unstandardized form in Figure 7 on page 160, and in standardized fonn in Figure 8 on page 161). Dwyer (1983) has suggested that a fully lagged model is most appropriate for modeling repeated time series data such as those in this study. However, because of high construct to construct covariation across time leaving little variation for the rest of the model, these types of models may be difficult to interpret. In addition, fully lagged SEM models typically have greater numbers of parameters to estimate, which may require sample sizes laiger than are available. The m ajor difference between the Fully Lagged and Distress Outcome models is the addition of follow-up Avoidance Coping (with the appropriate regression coefficients and correlations) in the former model. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 43 Change Scores Longitudinal Model Rather than using psychological distress at Time 1 as the outcome variable and controlling for Time 0 distress in the model, another approach to testing the relationships between stressful life events, coping styles, and psychological symptoms (distress) is to use change scores rather than controlling for baseline levels of dependent variables. Given the manner in which the Life Events construct is operationalized, however, a change score for this construct is not interpretable. In essence, the Life Events score at Time 1 already represents the new or “changed” Life Events from the previous 12 months. If an individual had 10 m ajor life events occur in 1989-1990 and then 10 more in 1990-1991, a Life Events Change score would be 0, which clearly does not adequately reflect the high level of stressors that this person had experienced. In fact, sustained high levels of stressors may be more damaging than a single year of high levels, not less damaging as a change score would imply. The Change Scores Model is shown in unstandardized form in Figure 9 on page 162. The Change Scores model is sim ilar to the Distress Outcome model except that the final outcome variable is change in Distress from Time 0 to T m e 1 rather than Distress at T m e 1 with T m e 0 Distress controlled by direct entry into the model. Also, the intermediate Avoidance Coping construct is modeled as change in Avoidance Coping from T m e 0 to T m e 1 rather than simply T m e 0 Avoidance Coping (in this respect the model is more like the Fully Lagged model). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 44 M ethodological Issues A number of potential confounds and methodological issues should be addressed in a study of coping-stress relationships. First, certain items on stressful life event scales are possibly indicators rather than causes of psychological symptoms (Dohrenwend et al., 1984). Although this is a serious criticism of stress research, it is more of a problem for the Daily Hassles based instruments than for the m ajor life events ones. For example, the Daily Hassles scale includes the items “you have had a fear of rejection” and “you have had trouble making decisions,” which Dohrenwend et a l.’s (1984) clinical psychologist raters deem to have high psychopathological content. One way to test for the effect of such a confound, if it exists, is to test similar models using the group of “m ore” versus “less” confounded measures (Lazarus, DeLongis, Folkman, & Gruen, 1985). Second, many previous prevention programs for adolescents have used as outcome variables behaviors classified as problem behaviors. These include drug use and delinquency (Pentz, 1994). In contrast, we view these as two of several modes of avoidance coping. The causal relationship between these coping methods and psychological outcomes are clarified in the current study by perfonning longitudinal analyses to sort out antecedents from outcomes. Third, a significant main effect found for a variable in a multivariate model does not mean, strictly speaking, that the variable has an effect on its Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 45 own—it means that it has an effect after other variables have been taken into account. It may have no effect in the absence of other variables (Rutter, 1985). This is more an issue of interpretation than one of study design or data analysis. In other words, if a main effect is found for avoidance coping in a multivariate model containing stressful life events, we do not know whether the use of avoidance coping in the absence of stressful life events would have an effect on distress. Some researchers (Holahan & Moos, 1991) have tried to clarify this point by examining whether the structure of the coping-distress relationship differs under conditions of low versus high stressors, but it is not directly addressed in this study. Im plications fo r H ealth Prom otion and Disease Prevention Previous prevention program evaluations for adolescents have focused either on reducing distress directly through primary prevention or througii programs targeting high risk youths (Pentz, 1994). High risk is defined by levels of stressors experienced, e.g., children whose parents were recently divorced. Alternatively, programs have focused on reducing single maladaptive coping behaviors (e.g., reducing drug use or delinquent behavior) studied not as coping behaviors but as problem-behavior outcomes (Pentz, Dwyer, M acKinnon, Flay, Hansen, Wang et al., 1989a; Perry & lessor, 1985). For example. Wills (1985) has suggested that substance use can constitute a coping mechanism—specifically, it can be used as an affect regulator, functioning either as a distraction from the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 46 problem or serving to enhance perform ance through increased arousal (depending on the individual, the situation, and the substance in question). No previous prevention program has looked at the mediator/m oderator distinction in detail nor looked at the interrelationships of multiple coping strategies simultaneously as a cluster. A finding in a longitudinal study that avoidance coping is a true mediator of the stressor-psychological symptom relationship would tend to support the use of prevention programs that attempt to reduce these behaviors. In addition, if the hypothesis that approach coping has a protective effect is supported, programs that focus on promoting adaptive coping strategies in adolescents, such as social (Pentz, 1985), life skills (Botvin et al., 1990), and resistance skills training (Hansen, Johnson, Flay, Graham, & Sobel, 1988; Tobler, 1986) will gain theoretical justification beyond that derived from evaluation of changes in skill levels. Since the true interrelationships of types of coping, stressors, and psychological distress are not known, it is difficult to design a theory-based coping intervention. However, without intervention testing, even if the theoretical relationships are more firm ly established, the relative efficacy of programs that reduce nonproductive coping compared to programs that promote productive coping is not known. In part this is because it is not known in this context whether it is easier to teach an adolescent to stop using an existing nonproductive coping strategy or to begin using a new productive strategy. This study evaluates Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 47 multiple types of coping strategies simultaneously in a longitudinal study to enable design of prevention programs that take into account risk factors and the use of maladaptive coping strategies. S t u d y H y p o t h e s e s The specific aim of this dissertation was to examine the structure of the stressor-coping-psychological distress relationship in adolescents with specific emphasis on the differentiation between adaptive and maladaptive coping strategies. Structural equation modeling was used as a technique to separate issues of measurement from those of conceptual structure. Specific hypotheses included those related to the direct adverse effects of stressful life events on adolescent psychological symptoms, those related to mediation of these effects by the use of maladaptive coping techniques, and those related to moderation or buffering of these effects by adaptive coping strategies. Secondary hypotheses included those related to gender and age. See Figure 1 on page 35 for graphic representations of direct, mediated, and moderated models of the stressor-coping-psychological distress relationship. This study offers new insight into the conceptual structure of coping styles and the coping process during adolescence. The findings will help intervention planners identify specific aspects of the coping process to target for modification by prevention programs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 48 Major Hypotheses Note that in some of the hypotheses below, mathematical expressions of the model parameter relationships implied by the hypothesis are given. Parameters SC, SD, and CD are the path coefficients between stressors and distress, stressors and avoidance coping, and avoidance coping and distress, respectively as shown in Figure 1 on page 35. Distress is used interchangeably with psychological symptoms. Avoid, Approach, Symptoms, JobStress, and SocialStress refer to mean levels of avoidance coping, approach coping, psychological symptoms, job-specific stressors, and social stressors. Subscripts specify that the parameters apply to only one of two groups in the tullowing pairs: H, L High versus low approach coping; O, Y Older versus younger adolescents; M , F M ale versus female gender. The parameter relationships are repeated in Table 1 on page 55. C onstruct/M easurem ent M odel Hypotheses Hvpothesis 1. Domains of Stressors Stressors experienced by adolescents were hypothesized to have distinct domains. Specific domains hypothesized to exist include social stressors, job- related stressors, and external environment stressors. The major developmental task of adolescence, identity fonnation, has several distinct domains and we hypothesized that specific stressors exist in each of these domains. Thus the stressful life event latent variable or construct was hypothesized to possess a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 49 distinct factor structure, representing the lower order constructs of the domain specific stressors— job, social, and environment—rather than a single or global stressor factor. Hypothesis 2. Domains of Coping Coping was hypothesized to be divided into several types, falling into four broad groups, rather than a single or global coping factor. These groups are cognitive approach coping, behavioral approach coping, cognitive avoidance coping, and behavioral avoidance coping. The coping response to potentially challenging or threatening internal or external stimuli can be either cognitive or behavioral in nature and furtherm ore either avoids the threat or in some way attempts to actively address either the problem or its psychological consequences. Avoidance coping construct was hypothesized to possess a distinct factor structure, representing the specific coping methods of aggression, substance use, individual relaxation, and cognitive restructuring. The approach coping construct was hypothesized to have factors representing physical exercise, calming behaviors, seeking social support, and decision making. Structural Hypotheses Hvpothesis 3. Stressors and Psvcholoeical Distress Stressful life events, as challenging and threatening events in persons lives, are likely to cause adverse psychological symptoms in some who experience them. Significant positive regression coefficients were therefore Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 50 hypothesized between Time 0 stressors and Time 1 psychological symptoms including anxiety, depression, and somatization. SD^ > 0 and SDp^ > 0. Hypothesis 4. Avoidance Coping and Psychological Distress As avoidance coping addresses neither the threatening problem directly, nor the intrapsychic responses to the problem, avoidance coping was conceptualized as a non-productive coping strategy. Significant positive regression coefficients between Time 0 avoidance coping and Time 1 psychological symptoms were therefore hypothesized. CD ^ > 0 and CDp^ > 0. Hypothesis 5. Avoidance Coping as a M ediator For some persons, some of the coping responses used will be avoidance coping. Because avoidance coping was hypothesized to be a non-productive coping strategy, it was hypothesized to mediate part of the relationship between stressful life events and psychological symptoms. Positive regression coefficients between Time 0 stressful life events and Time 0 avoidance coping, and between Time 0 avoidance coping and Time 1 psychological symptoms were therefore hypothesized. Hypothesis 4 is a necessary precondition to this hypothesis. In addition, SC^ > 0, SC ^ > 0, and the mediated effect SD=*'SC is significant in at least the low approach coping group, with the t statistic given in equation 1 on page 36. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 51 Hvpothesis 6. Approach Coping as a Stress Buffer Approach coping was conceptualized as a productive coping strategy that addresses either the specific problem or its emotional sequelae. The use of approach coping was hypothesized as protective against the adverse psychological effects of stressful life events. An interactive effect on one of the parameters relating stressful life events to psychological symptoms was hypothesized. SD l > SDpi and/or CD^ > CDpq. We expected to observe this effect on one or both of the regression coefficients whose dependent variable is psychological symptoms. That is, we predicted an attenuation of the effect of stressful life events on psychological distress in the group of adolescents who use more approach coping (smaller regression coefficients). We also expected the regression coefficient between avoidance coping and psychological symptoms to be smaller in the group with high approach coping skills. Finally, we expected this buffer effect to have a significant overall effect on psychological symptoms. That is, the mean levels of psychological symptoms would be significantly lower in the high approach coping group than in the low approach coping group. Symptoms^ > Symptom s^. No prediction about the effect of approach coping on the relationship between stressful life events and avoidance coping was made. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 52 Minor Hypotheses Construct-Level Hypotheses Hypothesis 7. O ther Approach Coping Effects We expected there would be subgroups of adolescents that use a more diverse set of coping strategies than others. In particular, we hypothesized that adolescents who use many coping strategies will use both approach and avoidance coping strategies. We therefore expected that when compared with the low approach coping group, the high approach coping group would also have higher mean levels of avoidance coping. Avoid^ > Avoid^. Sub-Group Hypotheses Hypothesis 8. Gender Hypotheses Hypothesis 8a. Gender and Psychological Symptoms Females are at higher risk for mood disorders as compared to males. We predicted that female reported greater levels of psychological symptoms as compared to males. Symptomsp > Sym ptom s^ Hypothesis 8b. Gender and Coping We expected that females reported higher levels of both m ajor types of coping, approach and avoidance coping. Avoidp > Avoid^ and Approachp > A pproach^ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 53 Hypothesis 8c: Gender and Stressor/Coping/Symptoms Relationships Notwithstanding the gender related differences mentioned, the theoretically meaningful structure of the relationship between stressful life events, approach and avoidance coping, and psychological symptoms was expected to be substantially the same between males and females. Hypothesis 9. Age-Related Hypotheses Hypothesis 9a. Age and Stressful Life Events Older adolescents are more likely to be employed than younger adolescents. In addition, they are m ore likely to have developed significant interpersonal or romantic relationships with their peers. We therefore expected older as compared to younger adolescents will report increased levels of job- related and social stressors. JobStresSg > JobStressy and SocialStressg > SocialStressy. Hypothesis 9b. Age and Coping Older youths are likely to have progressed further down the path of identity formation than younger youths. As part of this process they will have developed a laiger repertoire of a variety of cognitive, behavioral, and interpersonal skills. We therefore expected that older adolescents will report greater use of both approach and avoidance coping as compared to younger adolescents. Avoidg > Avoidy and Approachg > Approachy. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 54 Hypothesis 9c. Age and Stressor/Coping/Symptoms Relationships Greater numbers of stressful life events or a more diverse group of coping methods are not expected to change the fundamental relationships between the theoretical constructs. The theoretically derived structure of approach coping acting as a buffer of the adverse effects of stressful life events and of avoidance coping on psychological symptoms was hypothesized to be the same between younger and older adolescents. M e t h o d s Study Design This study was conducted using data from the Midwestern Prevention Project (M PP). The M PP is a longitudinal school- and community-based trial for the primary prevention of drug use in adolescents. The M PP is implemented in the communities that constitute the Standard Statistical M etropolitan Areas (SMSA) of Kansas City and Indianapolis. Within each SMSA a 2 x 2 factorial design is used, with intervention condition (prevention program or delayed program control) and grade of initial intervention (six or seven, representing the transition year from elementary school to either middle or junior high school) as independent variables. Indianapolis represents a three year lagged randomized experimental replication of the original quasi-experimental design in Kansas City. In Indianapolis, three successive cohorts of students are followed, representing a pure control cohort that was in sixth and seventh grades in the 1986-87 school Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 55 Table 1. Hypotheses and their Significant M odel Implications. Hypothesis Significant Parameter Relationship Expected Hypothesis 1 Factor structure of Stressor construct (multiple factors versus a single factor) Hypothesis 2 Factor structure o f Avoidance Coping, distinct from Approach Coping (two factors rather than one) Hypothesis 3 S D l > 0 and SD ^ > 0 Hypothesis 4 C D l > 0 and C D ^ > 0 Hypothesis 5 S C l > 0 and SC^ > 0 and Hypothesis 4 and mediated effect SD*SC is significant, with the t statistic given in equation (1) on page 36. Hypothesis 6 SD l > SDpj and CD ^ > C D ^ and DistressL > Distress^ Hypothesis 7 Avoidpj > Avoids Hypothesis 8a Distressp > D istress^ Hypothesis 8b Avoidp > A void^ and Approachp > A pproach^ Hypothesis 9a JobStresso > JobStressy and SocialStresso > SocialStressy Hypothesis 9b Avoido > Avoidy and Approachg > Approachy Note: Param eters SD, SC, and CD are the path coefficients between stressors and psychological symptoms (distress), stressors and avoidance coping, and avoidance coping and distress, respectively. Subscripts O: older, Y: younger, H: high approach coping, L: low approach coping. F: female, M: M ale. year with no schools receiving intervention, a mixed intervention/control cohort that was in sixth and seventh grades in the 1987-88 school year with schools randomized to either an intervention or control group, and a pure intervention Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 56 cohort that was in the sixth and seventh grades in the 1988-89 school year, with all schools receiving intervention. This study examined the Indianapolis students from all three cohorts. The stressful life events, coping, and psychological symptom questionnaire items were first introduced during the 1990-91 school year in the high school form, hence the baseline data for this study were collected from all three cohorts of students (9th only, 9 /10th, 10/11th grade) during the school year 1990-91. One year follow-up data was collected in 1991-92, in this year the three cohorts were in 9th/10th, lO th /llth , and 11th/12th grades. Grade cohorts, rather than individual students, were followed and surveyed. Individual students were identified and tracked by unique identification numbers. As this study was not examining program effects, all three cohorts were combined for analysis. D ata Collection Procedures A 20 page, approximately 35-45 minute self-report survey was administered by trained data collection staff. Students were surveyed in school during regular classroom hours. The survey was developed from prior research (Graham, Flay, Johnson, Hansen, Grossman, & Sobel, 1984) and included demographic, psychosocial, and drug use items. Students were identified only by code num ber and were assured their anonymity would be preserved. In addition to the survey, a biochemical measure of carbon monoxide (CO, a byproduct of cigarette and marijuana smoke) was administered to students Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 57 as a “bogus pipeline” to increase the accuracy of self-reported cigarette use (Bauman & Dent, 1982; Pechacek, Murray, Luepker, M ittelmark, Johnson, & Shutz, 1984). The use of CO as a pipeline refers to the collection, recording and analysis of CO to biochemically validate self-reported tobacco use. “Bogus” refers typically to the collection of the CO data with no intention of recording or analyzing CO as a measure of tobacco use. Several studies have shown that the accuracy of self-report of socially undesirable behavior is increased when subjects believed that an objective measure was also being used to assess the behavior (Bauman & Dent, 1982; Evans, Hansen, & M ittelmark, 1977). Immediately prior to the administration of the self-report questionnaire, a sample of expired air was obtained from each student and the concentration of CO was estimated using the M iniCO Indicator (Catalyst Research Corporation, Owings Mills, M D). Each student was instructed to inhale deeply, hold his/her breath for 10 seconds, and then exhale through a straw to blow up the balloon attached to the measurement device (Pentz, MacKinnon, Flay, Hansen, Johnson, & Dwyer, 1989c). In this study, CO levels were measured but not recorded. However, previous research has demonstrated CO/self-reported cigarette use correlations ranging from .50 to over .70 for 12th grade students (Pechacek et al., 1984). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 58 Table 2. Sample Characteristics Variable Entire Sample M issing Sample Analysis Sample N = 8036 N = 7 6 N = 401 % (SE) % fSE) % (SE) Ethnicity (% White) 73.8 (0.5) 76.3 (4.9) 76.8 (2.1) Gender (% Male) 48.8 (0.6) 46.1 (5.8) 49.4 (2.5) H igher SES' 55.9 (0.6) 64.5 (6,1) 58.1 (2.!% 9th Grade 44.3 (0.6) 44.7 (5.7) 42.4 (2.5) 10th Grade 33.2 (0.5) 34.2 (5.5) 33.4 (2.4) 11th Grade 22.4 (0.5) 21.1 (4.7) 24.2 (2.1) Note: “ H igher SES is the percentage with fathers in a professional or managerial occupation. No significant group differences existed between the Analysis Sample and either the Entire Sample or those with Missing data. Demographics are from the 1990-91 data collection. Sam ple Subjects were 9th, 10th and 11th grade high school students who were randomly sampled by classroom from 26 high schools in the metropolitan and suburban areas of Indianapolis between October 1990 and February 1991 and then randomly sampled again one year later. The “panel” used for this study is contained within the larger set of grade cohorts and consists of those individuals (identified by matching identification numbers) who received the questionnaire fonn with stress and coping items in both 1990-91 and 1991-92. The eligible sample from the grade cohorts was 8451 in the 1990-91 school year and 8311 the following year. Following elimination of duplicate identification numbers, the sample sizes were 8437 and 8289 respectively. O f the 8437 present in 1990-91, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 59 6945 or 82.3% were surveyed one year later. Four survey forms were randomly distributed to students, using sequential item sampling to minimize the length of any one form (Graham et al., 1984). Students who received a given form in the 1990-1991 school year did not necessarily receive the same form the following year. The sample responding to the form with the stress and coping items was 2389 in 1990-91 and 2115 the following year. O f the 2389 who responded to the appropriate sample at Time 0, 1946 or 81.5% were present at follow-up one year later. We would therefore expect approximately one fourth of these, or 487, to have received the same form both years. The actual number of students who responded to the fonn which included all o f the life events, coping, and psychological symptom variables in both years was 477. The sample used for the structural equation analyses in this study consisted of the 401 students with complete data from among the 477. Characteristics in 1990-91 of the analysis sample, the 8036 students present in 1990-91 who were not part of the analysis sample (8437-401=8036), and of the 76 students out of 477 that received the stress and coping survey but did not have com plete data are given in Tkble 2 on page 58. The 401 students did not differ significantly on the demographic variables from the 76 without complete data, nor did they differ from the students who received the other fonns. O f the 23% of the analysis sample that was non white, 22 % were African-American and 1 % were of other ethnicities. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 0 Measures The survey, adapted from previous research (Graham et al., 1984; Pentz et al., 1989a; Pentz et al., 1989c), included demographic characteristics, drug use items, and psychosocial items. Test-retest reliabilities of the m ajor items averaged .55 to .74 (Pentz, Pentz, & M acKinnon, 1988). The survey included 14 items measuring m ajor life events, 22 items measuring coping strategies, and 13 items measuring psychological distress. The life event items were adapted from the Life Events Checklist (Johnson & M cCutcheon, 1980). The item asks: “Below is a list of events that sometimes happen to people your age. For each of the events check if the event happened to you in the last 12 M O N TH S.” The response choices were yes and no. After pilot testing, items with 0% response rates were eliminated and the wording of some items modified to increase their relevance to adolescents, e.g. “increased arguments with parents” instead of “increased arguments with spouse.” Coping strategies were measured with twenty items from Wills’ Coping Inventory: three cognitive coping items; three aggression items; two physical exercise items; three substance use items; two social entertainment items; three individual relaxation items; one decision making item; one parental support item; one peer social support item; one prayer item (Wills, 1985). TVvo additional items were added to the scale. The psychological distress items were adapted from the SCL-90-R (Derogatis, 1975). Preliminary work with a sim ilar data set from the Midwestern Prevention Project (Broder, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 1 Pentz, & Chou, under review) used a somewhat different conceptualization of coping. However, maladaptive coping strategies were found to include aggression (e.g. I get mad at people, I do something bad or cause trouble), substance use (I smoke a cigarette, I take pills, I drink alcohol or use drugs), cognitive restructuring (e.g. I try not to think about it, I tell myself the problem is not worth getting upset about, and I just wait and hope that things will get better with time), and social entertainment (I hang out with friends, I go to or have a party). M aladaptive coping is essentially congruous with avoidance coping. Approach coping strategies included physical exercise, talking with peers or parents and getting information, and calming behaviors (I try deep breathing, I try to calm myself). The entire questionnaire is appended to this proposal. Coping and psychological distress items were measured using a four point Likert-type scale. The response categories ranged from “0 = almost never” to “4 = almost always.” For the coping items, the scale reflected the frequency of use of each coping strategy when confronted with a problem. The psychological distress items asked how frequently each of the experiences or feelings occurred. For the life event items, students were asked whether they had experienced an event in the last 12 months. The questionnaire used during the 1991-92 school year appears in the appendix. Questionnaire item scores were standardized and then means were taken to obtain sub-indices for the different coping and psychological distress constructs. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 2 The standardized scores were summed to obtain the sub-indices for the stressor constructs. It is these indices that were used as the input to the structural equation modeling procedure. Because they were derived directly from the questionnaire items, they are referred to at times as “measured variables” for the purposes of distinguishing between the SEM inputs and those latent constructs that are unmeasured. Statistical Power ly p e I error is the error of rejecting a true hypothesis. In practice, we choose an a (significance level) or Type 1 error level which is acceptable based on the type of hypothesis being tested, with the overlying convention that a = .05 is the maximum acceptable a. to choose. This means that if the true hypothesis is the null hypothesis, that there is no effect to observe, we will stand a .05 or 5% chance of incorrectly rejecting the null hypothesis. In contrast, ly p e II errors, or beta errors, are errors of accepting a false hypothesis. W hen designing a study to detect a program effect, and assuming that the program effect is present, the false hypothesis that we do not want to accept is the null hypothesis. In practice, we make our study large enough to reduce our chance of missing a true effect to an acceptable level. Power, or (1 - beta), gives the probability of rejecting the null hypothesis when it is false. It is the probability that the test will detect an effect in a sample when a true effect is present in the population from which the sample is derived. Commonly accepted levels of power are .80 or .90, meaning that if Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 63 the null hypothesis false in the population (therefore a significant effect exists in the population), we have a 80% (or 90%) chance of rejecting the null hypothesis with a study sample of the given size. As this study is a secondary data analysis, the sample size is fixed in the power calculations. Rather than making assumptions about effect size to estimate the sample size necessary to obtain specific statistical power (/?), we will make assumptions about statistical power to show with what confidence we can expect to detect effect sizes with varying sample sizes. Using a correlational model, with a significance level of a = .0 5 , the conservative assumption of a two-tailed test, and a sample size of 443, sample correlations (r) of greater than 0.093 are significant. Therefore, if the population correlation is greater than .13 we would have statistical power of .80. If we require power of .90, then the population correlation must be greater than .15. In a regression framework we can express the effect size as an incremental R' obtained when the variables of interest are entered into the regression model. A sample size of 443 implies that power of .81 can be obtained with incremental of only .017. All results were obtained from the POWER program for DOS (Borenstein & Cohen, 1988). Aiken and West (1991 pp. 156-165) provide a more sophisticated discussion of statistical power analysis of multiple regression interaction models. Power calculations for regression models such as that given in equation (2) on Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 64 page 38 group together all first order effects and refer to these as the set M (“m ain”) effects. In a simplified model the single interaction term is called the I (“interaction”) effect. They define the following terms for use in the power analyses (Aiken & West, 1991 p. 157): ry-M- The squared multiple correlation resulting from prediction of Y by the set M of main effects only, r y MI- The squared multiple correlation resulting from combined prediction by the main and interactive sets of regression terms. /"Y ( i • M )’ The gain in squared multiple correlation due to the addition of the interaction terms to an equation already containing the main effects terms. It is the proportion of total variance accounted for by set I, over and above set M. I'x-z- The interpredictor correlation (e.g. the correlation between approach coping and avoidance coping). f~\ Effect size for the interaction set I over and above the main effects set M , where effect size is defined as the proportion of systematic variance accounted for by the effect relative to the unexplained variance in the outcome variable. Several things are important to note. First, the squared partial correlation and the effect size are similar, both contain as their numerators the squared semipartial correlation, that is the proportion of total variance explained by the interaction Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 65 term or terms over and above the main effect terms. However, the denominator for the effect size is the residual variance after entry of both the main and interaction effects into the model; the denominator for the squared semipartial correlation is the residual variance only after entry of the main effects. For small effects sizes they are approximately equal. Second, it is the effect size rather than the percentage of variance explained by the interaction term (squared semipartial correlation) that is important in determining statistical power. In other words, models to detect interaction effects will be more powerful if the main effects explain much of the variance before entry of the interaction tenus. Aiken and West’s Tkble 8.5 (Aiken & West, 1991 p. 164) presents sample size requirements to obtain a power of .80 with a = .05 for different values of / " y - m (percentage of variance explained by the main effects), different effect sizes, and different assumptions about the reliability of the measures. Assuming no measurement error and a small effect size / - = .02, a sample size of 392 is required for power of .80 (at a = .0 5 ). Preliminary results indicated that for these data ry-M is approximately .15 and the gain in variance explained by the addition of the interaction tenns is approximately .03, yielding an effect size of about .035. From Aiken and West’s table it can be seen that sample size requirements have a close to linear relation to effect size, so for these data, given perfectly reliable measures, a power of .80 with « = .0 5 would require a sample size of 224. Unfortunately, if the measures have reliabilities of .80, then the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 6 6 sample size requirements approximately double. In fact, they present sample size requirements of approximately 700 to 800 for power of .80, given a = .05, measure reliability .80, effect size of .02, interpredictor correlations of between 0 and .50 and variance explained by main effects of between 0 and .20. Assuming that the effect size that exists in the population sampled for this study is .035 yields sample size requirements of 400 to 457. Thus it appears that the sample size in this study is adequate to detect the hypothesized interaction effect in 80% of the samples taken from this population. Data Analysis Statistical analysis took place in several stages: (1) data cleaning, (2) preliminary univariate analyses and scale construction, and (3) substantive analyses of the primary and secondary research hypotheses. Data Cleaning Frequencies of all study variables were computed in order to identify and correct errors in coding or data entry. Descriptive univariate statistics including minimum, maximum, skewness, and kurtosis were used to identify potential problems with individual items. Items with moderate levels of skewness and kurtosis were retained for analysis, with the understanding that composite scales based on such items may also be quite non-normal in distribution and may make subsequent structural equation models more unstable. After the data were cleaned. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 67 subjects with complete data on the substantive variables of interest were included in the analytic sample. Sample Characteristics Data analysis included simple t-tests and Mantel-Haenzel tests to establish that the study sample accurately represents the population by comparing the final sample with the samples before students with missing data were eliminated. Variables used for these comparisons will include demographic variables and study variables where possible. Scale Construction Subscales for the stressful life events were formed by simple sums of subsets of the standardized items specific for each of the hypothesized stressor domains. Subscales for coping and psychological distress items were the means of the appropriate sets of standardized items. Exploratory and confirmatory factor analyses guided by substantive theory were conducted to determine both the items to be included in the subscales and the subscales to be included in structural equation modeling with both single-group (full sample) and multiple-group approaches. Constructs for the coping and psychological distress subscales were validated using a common factor analysis with promax (oblique) rotation (SAS Institute Inc., 1988). Items with loadings below .30 on all factors and those that could not be assigned cleanly to a single factor were dropped following Comfrey (1978). Factors with low internal reliability were also be dropped unless strong Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 68 substantive reasons for their inclusion existed. Cronbach’s a ’s are reported for each subscale in Tkble 8 on page 90, however because the life events are expected to be independent (i.e. mutually random events), intercorrelations may not be high and a may not be an appropriate measure of reliability (Newcomb et al., 1981). Tests o f th e M ain H ypotheses Structural Equation M odeling Overview The major objective of the structural equation modeling process in this study was to test the hypotheses concerning the differential effects of approach and avoidance coping on the stressor-psychological distress relationship, as well as to test the hypotheses concerning this relationship itself. Secondary objectives included testing hypotheses related to gender and age. Structural & Measurement M odel vii VI V3 V12 Stressor) Distress! V4 V13 Avoidanc Coping J V9 VIO V8 F ig u re 2. Structural and M easurement M odel Structural equation modeling consists of a method of specifying the structure of an observed phenomenon in terms of a set of observed variables. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 69 Two models are specified: both a measurement model, that is, the relationship of the observed variables to the hypothesized or latent variables; and a structural model specifying relationships among the latent variables (Joreskog & Sorbom, 1988). This dual specification, as shown in Figure 2 on page 68, allows the separation of issues of validities and reliabilities (measurement issues) from issues of causal inference and unexplained variance (structural issues). The structural relationships are estimated while controlling for measurement error (Kenny, 1979). Furthermore, multiple hypotheses about the structural relationships may be tested simultaneously. In particular, structural equation modeling may control for some of the measurement issues involving coping raised by Lazarus, Averill, and Opton (Lazarus, Averill, & Opton, 1974) and by Wills (1986) and as well as measurement issues involving life stress raised by several groups (Dohrenwend et a l., 1984; Lazams et al., 1985; Newcomb, Huba, & Bentler, 1986). In addition, structural equation modeling allows us to test for complex interactions between approach coping and the other parameters of interest. In particular, to test for a classic stress buffer effect of approach coping, we would expect to see significant differences between models for a group of students with high levels of approach coping versus low levels of approach coping. By using the multiple groups approach to structural equation modeling, we can not only observe these differences if present, but we can also test them for statistical significance. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 70 The models presented can be evaluated on several levels. Before considering individual parameters of interest, the overall fit of the model must be assessed. EQS (Bentler, 1989) can use maximum likelihood estimation to provide an overall x ' goodness-of-fit test. Assuming multivariate normality, the x^ test has the statistical properties normally associated with the x^ statistic. This tests the difference between a given model and a completely saturated model that perfectly represents the data. That is, a statistically significant x^ value allows one to reject the current model in favor of one with fewer restrictions which more accurately models the data. However, because the x^ test is directly related to sample size, a model that in a practical sense adequately represents the data may have a significant x " if the sample size is large (i.e. small residual covariances are statistically, but not practically significant at large sample sizes). One method of handling this situation is to test the model using random subsamples small enough that X' is interpretable. However, models can become unstable at smaller sample sizes (Comfrey, 1978). M ore commonly, fit indices other than x^ are used to evaluate the overall goodness-of-fit of the model. Bentler and Bonett (1980) proposed that this alternative model be the null model rather than the perfect model used in the x " test (Bentler, 1990a; Bentler, 1990b; Bentler & Bonett, 1980). They proposed several goodness of fit indices. The first, the non-nomied fit index (NNFI), assesses the proportion of covariance of observed variables explained by the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 71 model compared to the null model. According to Widaman (1985), it is a relative measure of the covariance because the fit of the model is assessed relative to the degrees of freedom for the model. The second measure, the normed fit index (NFI), is an absolute measure of the covariance because it is not adjusted for the degrees of freedom. Finally, the comparative fit index (CPI), has the advantage that it reflects fit relatively well at all sample sizes, rather than underestimating fit. Both CFI and N FI range from 0 to 1, with 0.9 typically being considered an empirically adequate fit (Bentler & Bonett, 1980). Finally, just as the fit of individual models can be assessed by comparing them to the null model, competing models can be compared if they are hierarchically nested (Bentler, 1990a; Widaman, 1985) by examining the difference between the two fit indices. This can be done whether the N FI, NN FI, CFI or another fit index is being used to assess fit, but if a difference of x^’ s is used, we can statistically assess the significance of the two models. A statistically significant difference would mean that the less restrictive model better represents the data observed. A non significant difference would mean that we could not reject the restrictive model and would tend to favor it for reasons of parsimony (fewer estimated parameters make it preferable). O f course, it is the param eter estimates that are of interest in any given model. EQS provides standard errors for each parameter estimate. Dividing the parameter estimate by its standard error yields a z-ratio. Absolute values of the z- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 72 ratio greater than 1.96 are statistically significant at the a = .05 level, implying that we can reject the null hypothesis that the parameter in question is equal to zero. EQS provides the p-value associated with each parameter. To improve the model fit, correlated errors suggested by substantive and methodological considerations were added to the basic models guided by the Lagrange M ultiplier test (Chou & Bentler, 1990) in EQS. The added correlated errors decrease the difference between the observed covariance matrix and the covariance matrix implied by the model and therefore improve the fit of the model. Hierarchically nested models were compared by examining the difference between goodness-of-fit x ‘ statistic (Bentler, 1990a; Widaman, 1985). All structural equation modeling analyses were conducted on a moment m atrix which augments the infonnation available in a covariance matrix with information about the means of measured variables and allows testing for differences on mean levels of structural constructs across groups. Measurement Model The first stage of model fonnation was to establish the measurement model. Adequate measurement models constitute the building blocks for the investigation of substantively meaningful causal models. Both the structural and measurement models are shown in Figure 2 on page 68, and one portion of the measurement model is shown in more detail in Figure 3 on page 73. In Figure 2, the structural model consists of the three latent variables in the circles (stressors. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 73 avoidance coping, and psychological distress) and the arrows connecting them, whereas the measurement model consists of the latent variables and the arrows that connect them to the measured variables Vs in the boxes. As shown in Figure 3 on page 73, each measured variable has an associated error and the latent constructs each have a disturbance, similar to the errors for the measured variable. Measiurement Model in Detail Disturbance Avoidanc Coping Substance Use Social Avoidance Error Error Error Measurement model construction was done cross- sectionally fo r the entire analysis sample and was driven by the theoretical assumptions about intra- and inter-construct relationships. In particular. F ig u re 3. M easurem ent M odel in Detail "’“ 'j” latent constructs are assumed to be only indicators for the construct predicted by theory. That is, except in a single instance, indicators were only allowed to load on factors suggested by substantive theory and face validity (i.e., approach m ethods of coping will only allowed to load on the approach coping factor, not on avoidance coping, stressors or psychological distress). All the major constructs—stressors, approach coping, avoidance coping, and psychological distress—were expected to be correlated with one another. M oment matrices and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 74 maximum likelihood solutions were used as the data are sufficiently normal in distribution (Pentz & Chou, 1994). M odel reduction was performed, with unsatisfactory indicators being removed, as long as the requirement of three indicators per construct is not violated (Bentler & Chou, 1987). If the initial measurement model fit is inadequate, with significance less than .05, or nonned fit index (NFI) (Bentler & Bonett, 1980) or comparative fit index (CFI) (Bentler, 1990a) less then .9, then meaningful parameters included correlated errors can be added to the model guided by modification indices such as the Lagrange M ultiplier (LM) test (Bentler, 1989; Chou & Bentler, 1990; Joreskog & Sorbom, 1988). If the measurement models derived for each wave of data show substantially the same structure, a common measurement model can be used for the longitudinal structural model. The testing of the measurement model, as well as the scale construction process described above, will constitute tests of hypotheses 1 and 2. The next step in the model testing process, for each of the five structural equation models tested, is the testing of multiple-group models. This step begins with the testing of single-group models. Single Group Structural Model Once the measurement model was confirmed, the first structural models were tested. Models were fit separately for the low and high approach coping groups. These models used data from both waves to test relationships between stressors at lim e 0 and coping and psychological symptoms at Time 1. If the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 75 initial single group structural model had inadequate fit, free parameters were added based on substantive and practical measurement considerations and guided by the LM test procedure. The danger in this procedure is that error correlations added may be without theoretical justification and as the error correlations to be added are not specified in advance, the process of model modification to improve model fit may be considered to be data- rather than hypothesis-driven. Bentler and Chou (1992) suggest some solutions. Since the model parameters of interest can and are specified a priori, by examining how they change during the process of model modification we can try determine whether the model fitting process itself is fundamentally changing estimates of the parameters of interest. Specific methods to test the changes include computing correlation coefficients between the set of common parameters of the basic model and each of the modified models. T-tests can also be used to evaluate differences on mean parameter estimates of models before and after modification. In this study, con-elations between sets of parameters in the initial and final models are used as a confimiation that the model modification process did not substantively change the parameters of interest. W here the correlations suggested that the set of parameters did change substantially, the parameters were examined for outliers to determine if the lack of agreement between initial and final param eter sets reflected differences in one or a small number of parameters rather than differences throughout the parameter set. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 76 Multiple Group Model The two groups, high and low approach coping (based on a median split) were then combined. EQS does not require a dummy group variable, only that the groups be specified separately in the model specification section. Then, in a succession of nested models, equality constraints (that is, requirements that specific parameters are equal across groups) of various types were imposed, model fit was examined, and if the model fit deteriorated significantly, individual equality constraints were released guided by LM or other tests until a model with adequate fit was obtained. The process was then repeated with the next group of equality constraints, using the final model from the previous iteration as the base model for the next (Pentz & Chou, 1994). In more concrete terms, call the first multiple group model MO. Adding equality constraints on the intercepts of measured variables results in model M l. If the difference between the x ‘ of MO and the x" of M l is significant, then some of the equality constraints must be released (guided by the LM or another test) until model M l* is obtained, whose fit does not differ significantly from MO. Then, using M l* as the comparison model, equality constraints on factor loadings are added, yielding model M 2. If x^ test of M 2-M 1* is significant, release constraints on M2 until M 2* is obtained, where the x ' test of M 2*-M 1* is non-significant. The relationship of M 2* to MO represents the relative measurement invariance that exists across the two groups, adolescents with high Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 77 and low levels of approach coping. As each successive model is nested in the previous one we may have total or partial measurement invariance across groups. Finally, the Predictions of Multiple Group Analysis High vs. Low Approach Coping • Distress “ lower in high approach coping group. • Avoidance Coping - higher in high approach coping group. • All regression paths expected to be lower in high approach coping group. Avoidanc Coping s u b s t a n t i v e l y i n t e r e s t i n g step, m o d i f y M 2 * to c o n s t r a i n f a c t o r variances, covariances. F ig u re 4. H ypothesized R elationships B etw een regression weights Constructs equal across groups, yielding M 3. The test for an interaction or stress buffer effect is whether o r not one or more of the equality constraints on regression weights between stressors and avoidance coping, stressors and psychological symptoms, or avoidance coping and psychological symptoms are unreasonable. Table 1 on page 55 summarizes the significant param eter relationships expected based on specific hypotheses. The parameters SD, SC, and CD are listed in the caption of Tkble I and are labelled graphically in Figure 1 on page 35. The m ajor hypotheses 3 through 6 are also summarized in Figure 4 on page 77. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 78 Regression Methods Three sets of regression analyses were performed to partially replicate the structural equation modeling analyses, though without correcting for measurement error. For all regressions, standardized indices were used. The Life Events index was the sum of the standardized life events measured variables. The Avoidance Coping, Distress, and Approach Coping indices were all means of their respective sets of variables. The first set of regression analyses replicated the Distress Outcome structural equation model, with Distress at Time 1 as the outcome variable and Time 0 Life Events, Time 1 Life Events, and Avoidance Coping as predictor variables. These analyses were conducted separately for the Low and High Approach Coping groups to provide preliminary indications of whether an interaction effect was likely to exist. Differences in regression coefficients between the two groups would be suggestive of such interaction effects. These analyses also provided population-based estimates of the amount of the variance explained by the models. The second set of regression models focused on an analysis of hypothesized mediator effects. Avoidance Coping was hypothesized to partially mediate the effects of Life Events on Distress. In eacii of the two Approach Coping groups, two regression models were estimated. These models corresponded to the model in the bottom half of Figure 1 on page 35. In the first Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 79 of these, Time 1 Distress was the dependent variable and Time 0 Life Events and Time 0 Avoidance Coping were the independent variables. Parameter estimates SD and CD are obtained from this model, along with their standard errors. In the second model in this set. Time 0 Avoidance Coping was the dependent variable and Time 0 Life Events was the predictor variable. This yielded the SC parameter and its standard error. M ediator effects were assessed by examining the mediated and direct effects (SC*CD and SD) and computing the z-ratio for the mediated effect as shown in equation (1) on page 36. The final regression model is an interaction model. The full analysis sample is used in a single model with Time 1 Distress as the outcome variable. M ultiple regression was performed with forced entry of the first order variables (main effects); Time 0 Life Events, Time 1 Life Events, Time 0 Avoidance Coping, Time 0 Approach Coping, and Time 0 Distress. Interactions were tested by adding cross-product terms to the regression model after the main effects. R e s u l t s Descriptive Statistics Observed Variable Characteristics Means, standard deviations, skewness, and kurtosis for each of the twelve measured variables (actually sub-indices) consisting of three measured indicators for each of four latent factors at each of the two times are given separately for the high adaptive coping group (lower triangle) and low adaptive coping group Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 80 (upper triangle) in Tàble 15 on page 156. Tkble 15 also contains the within-group correlations for the 24 variables. The three stressor variables are identified by having names ending in Str, so that SocStr refers to Social Stressors whereas Social refers to Social Coping. In general variables were normally distributed. Two of the stressor variables were slightly skewed and kurtotic in both groups: External Stressors were skewed and kurtotic in both groups at both waves, Job Stressors were skewed and kurtotic in both groups in 1990-91 and skewed only in 1991-92. Depression and Somatic symptoms were also slightly skewed and kurtotic in both groups: Depression was skewed and kurtotic in both groups at both waves, Somatic symptoms was skewed and kurtotic in both groups in 1991- 92, skewed only in the high adaptive coping group in 1990-91. Do Drugs Coping was skewed and kurtotic in both groups at both waves. Antisocial Coping was slightly skewed and kurtotic in 1991-92 in the low adaptive coping group and kurtotic in 1991-92 in the high adaptive coping group. However, the variables were "relatively" normal, since the most highly skewed and kurtotic variable, Do Drugs Coping was maximally skewed and kurtotic in the Low Adaptive Coping Group in 1990-91 (2.37 and 5.16, respectively). Significant group differences existed for Antisocial and Social Coping at both waves, with the low adaptive coping group having lower means (Antisocial Coping: -.16 and -.08 versus .17 and .10 for the high adaptive coping group in 1990-91 and 1991-92 respectively. Social Coping: -.18 and -.10 versus .17 and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 81 .09). Anxiety was significantly lower in the low adaptive coping group in 1991- 92 (-.06 versus .09). As expected, all three adaptive coping indices were significantly lower in the low adaptive coping group. Factor Characteristics Means, standard deviations, skewness, and kurtosis for each of the main indices (factors) at each of the two times as well as for the two change scores (Avoidance Coping and Psychological Distress) are given separately for the low adaptive coping group and high adaptive coping group in Tkble 3. The factors and their abbreviations include: Stressors (Stress), Avoidance Coping (NegCop), Psychological Distress (Distress), and Approach Coping (PosCop). Subscripts indicate time of measurement (Time 0 is 1990-91, Time 1 is 1991-92) or D for change scores. Table 3 also contains within-group correlations between the factors. Correlations whose absolute value is greater than .14 are significant at p < .05, I r I >.18 is significant at p < .01, and j r | >.23 is significant at p < .001. The low adaptive coping group had lower mean levels of Avoidance Coping at Time 0 ( p < .0001) and Time 1 ( p < .01) and of Approach Coping at Time 0 and Time 1 (p < .0001, as expected since Approach Coping at Time 0 was used to define the two groups). Lower mean levels of Avoidance Coping in the Low Approach Coping group confirms Hypothesis 7. The low adaptive coping group had marginally higher levels of change in Avoidance Coping and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 82 T ab le 3. Descriptive Statistics and Correlations o f Factors by Adaptive Coping Group Adaptive Coping High (N = 206) Variable 1 2 3 4 5 6 7 8 9 10 Mean -.20 .14 .04 .45 -.29 .08 .07 .25 -.05 .02 SO 4.24 .66 .67 .40 4.67 .62 .73 .60 .64 .61 Skew 1.36 .85 .72 -.67 1.05 .65 1.20 .27 -.32 .26 Kurt 2.17 .88 .13 -.57 1.23 1.03 3.29 -.40 1.75 2.06 1. Stresso -.085.19 .61 -.21 .18 .17 -.03 .22 .21 .09 .11 .03 -.08 2. NegCopo"* -.13 .51 1.74 3.81 .15 .39 .20 .07 .48 .19 .03 -.54 -.21 3. Distresso+ .06 .60 .96 .64 .26 .41 .29 .30 .30 .62 .01 -.11 -.38 4. PosCopo"* -.49 .31 1.15 1.05-.14 .25-.08 .04 .13 .17 .27 -.08 -.11 5. Stress, .02 5.35 1.05 1.12 .32 .17 .23 -.05 .20 .45 .00 .12 .20 6. NegCop,** -.07 .50 1.35 2.68 .07 .4 9 .2 7 .13 .18 .32 .30 .48 .04 7. Distress, -.03 .64 1.18 1.59 .17 .31 .61 -.00 .21 .29 .02 .11 .49 8. PosCop,*** -.22 .55 .27 .0 2-.03-.02-.00 .25 -.12 .22 -.01 .26 .02 9. NegCopo+ .06 .51 .30 2.89-.09 -.55-.16 -.13 -.01 .46 -.04 .23 .25 10. Distress^ .03 .54 .35 4.05-.08 -.08-.36 .09 .01 .05 .52 -.01 .12 Adaptive Coping Low (N = 195) Note: = p<AQ, * — /? < .0 5 , ’* = /7 < .0 1 , * * * — /?< .001 for group differences. | r | >.14 is significant at p < .05, |r| >.18 is significant at p < .01, I r I >.23 is significant at p < .001 for both groups. Subscripts: 0 indicates data are from 1990-91 school year (Time 0), 1 indicates data are from 1991-92 school year (Time 1), and D indicates differences or change scores (1991-92 minus 1990-91). marginally higher levels of Psychological Distress at Time 0 than did the high adaptive coping group (p < .10). Correlations between factors for the low adaptive coping group are shown in the lower triangle in Tkble 3, for the high adaptive coping group in the upper triangle. To provide protection against an inflated Type I error rate due to multiple statistical tests, a levels more stringent than the normal .05 can be used. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 83 Of the 49 correlations that are statistically significant p < .05, over 80% are significant at p < .01, and approximately two-thirds at /7 < .0 0 1 , so that even a more stringent a would not substantially alter the significance of the observed bivariate associations. S am ple G roup Equivalence A pproach C ooing G ro u p Equivalence Demographics for the low and high approach coping groups are provided in Table 4 on page 83. Statistical tests (two-tailed t tests) were conducted to T able 4. Demographics by Approach Coping Group Variable Low Approach Coping Hieh Approach Coping Years of Data Collection 1990-91 and 1991-92 N== 195 N = 206 % (SE) % (SE) 9th Grade 46.2 (3.6) 38.4 (3.4) 10th Grade 29.7 (3.3) 36.9 (3.4) 11th Grade 24.1 (3 1) 24.3 (3.0) Gender (Male) 50.8 (3.6) 48.1 (3.5) Ethnicity (White) 76.4 (3.0) 77.2 (2.9) H igher SES" 52.4 (4J) 62.8+ (3.7) Note: "Higher SES is the percentage with fathers in a professional or managerial occupation. = p < .1 0 for two-tailed tests o f group differences. Demographics are from the 1990-91 data collection. examine between group differences on mean levels of demographic variables. In general the low and high adaptive coping group were demographically similar. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 84 The high adaptive coping group were of marginally higher socioeconomic status—63% higher SES versus 52% for the low adaptive coping group (p<.10). Although not statistically significant, there were some differences in grade distribution between the two groups. There were more 9th graders (46%) and fewer 10th graders (30%) in the low adaptive coping group than in the high adaptive coping group (38% 9th and 37% 10th). G ender an d G rad e G roup Equivalence Demographic variables measured in 1990-91 for groupings by gender and by grade (9th versus 10th/11th grades) are given in Tkble 5 on page 84. There T able 5. Demographics by Gender and Grade Variable Female M ale % (SE) % (SE) 9th Grade 46.3 (3.5) 38.4 (3.5) 10th Grade 32.5 (3.3) 34.3 (3.4) 11th Grade 21.2 (2.9) 27.3 (3.2) Ethnicity (White) 74.4 (3.1) 79.3 (2.9) Higher SES" 55.2 (3.9) 61.2 (3.9) Variable 10th o r 11th Grade 9th Grade Ethnicity (White) 77.9 (2.7) 75.3 CE3) H igher SES" 58.2 (3.6) 57.9 (4.3) Gender (Male) 52.8 (3.3) 44.7 (3.8) Note: “ Higher SES is the percentage with fathers in a professional or managerial occupation. ^ — /?< .10 for two-tailed tests of group differences. Demographics are from the 1990-91 data collection. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 85 were no statistically significant mean demographic differences between those students in a lower grade (9th) versus those in higher grades (10th and 11th). M ales were maiginally older than females (average grade 9.9 for males versus 9.8 for females, p < .10). T able 6. Stress, Coping, and Distress Factor Means by Gender and Grade Bv Gender By Grade Factor Female M ale 10th/11th 9th Stressorg .482 -.7 8 3 " -.415 .238 Avoidance Copingg .022 -.004 .010 .009 Distressg .138 -.154"* .017 -.037 Approach Copingg .006 -.023 .015 -.039 Stressor; .037 -.326 .307 -.753 Avoidance Coping, .022 -.005 .037 -.030 Distress, .157 -.123"* .037 -.006 Approach Coping, .025 .019 .024 .019 Avoidance Coping^ -.000 -.002 .027 -.038 Distress^ .019 .031 .020 .031 Note: * = p<.05, ’* = p<.Ol, * * * = p < .0001, for two-tailed tests for group differences either between males and females or between 10th/11th and 9th graders. Subscripts: 0 indicates data are from 1990-91 school year (Time 0), 1 indicates data are from 1991-92 school year (Time 1), and D indicates differences or change scores (1991-92 minus 1990-91). Several factor means were significantly different by gender. Table 6 on page 85 gives stressor, coping, and distress factor means by gender and by lower versus higher grade. As hypothesized (Hypothesis 8a), females had higher mean levels of psychological distress symptoms at both times (.138 and .157 versus - Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 86 .154 and -.123 in 1990-91 and 1991-92 respectively). Consistent with Hypothesis 8b, mean levels of Avoidance Coping were higher in females than males (.022 and .022 versus -.004 and -.005 in 1990-91 and 1991-92 respectively) although the difference was not statistically significant. M ean levels of Approach Coping were also higher in females (.006 and .025 versus -.023 and .019) although once again these differences were not statistically significant. The only statistically significant difference on factor means by age was that the older students had significantly higher mean levels of stressors than the younger students. This is consistent with Hypothesis 9a. Hypothesis 9b refers to specific measured variables, but as shown in Ihble 7 on page 87, although Job Stressors were higher at both waves in the older students, there were no significant differences between older and younger students on Social Stressors. The only other significant difference on Life Events variables was lower levels of External Stressors in the older group at Time 0. This is likely because in 1990, the younger group was in 9th grade and had just changed to high school, so that one of the components of external stressors, “Changed to a new school,” was reflecting this normal transition of schools rather than, for example, a change between two high schools. Finally, although the mean levels of Avoidance and Approach Coping were higher in the older students than in the younger at both waves, none of the differences were statistically significant, so Hypothesis 9b could not be proved. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 7. Measured Variable Means by Grade Variable 10th/11th 9th Extstro*** -0.502 0.464 Jobstrg*** 0.296 -0.408 Socstrg -0.209 0.172 Antisog 0.048 -0.049 Dodrugo -0.001 0.047 Socialo -0.016 0.029 Anxietg 0.029 -0.053 Depresg 0.008 -0.034 Somatig 0.014 -0.025 Exercig -0.039 0.035 Getinfg 0.047 -0.073 Relaxg 0.037 -0.075 Extstrj -0.202 0.095 J o b s tr/" 0.461 -0.691 Socstrj 0.049 -0.157 Antisoj 0.026 -0.003 Dodrug; 0.043 -0.027 Social, 0.042 -0.059 Anxiet, 0.038 -0.009 Depres, 0.048 -0.029 Somati, 0.026 0.018 Exercij -0.022 0.076 Getinf] 0.062 -0.018 Relax, 0.034 -0.001 Note: ’ = p<.05, " = p<.01. p<.0001. two-tailed tests for group differences between 10th/11th and 9th graders. Subscripts: 0 indicates data are from 1990-91 school year (Time 0), 1 indicates data are from 1991-92 school year (Time 1). 87 M easurem ent M odel V alidation Factor Analysis an d Scale C onstruction Constructs for the coping and p sy ch o lo g ical d istress subscales were validated using a common factor analysis with promax (oblique) rotation (SAS Institute Inc., 1988). Based on exploratory factor analysis, the life events scale was divided into three subscales, representing stressors related to social factors, employment, and external factors. Two items, “parents away more often” and “failed a grade” were dropped because they had loadings below .30 on all factors. A three item internal stressors subscale was dropped because its most representative Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. item, “quit school” could not be expected to be fairly represented in a school-based sample and because the internal reliability for this scale was relatively low, particularly in 1991-92 (« = .4 6 versus .67, .64 and .52 for the other stressor subscales in 1991-92). In general, the sub-indices had adequate reliabilities, show in Ikble 8 on page 90. Five of the twelve sub-indices had reliabilities over .70. As expected, the stressor sub-indices had low reliabilities, particularly in 1990-91, ranging from .45 to .67. Without the stressor sub indices, five of the nine sub-indices remaining had reliabilities over .70. There is a strong theoretical basis for the overall use of a life events score as a measure of stressors and for the use of domain-specific subscales (Newcomb et al., 1981; Rahe, 1974; Skinner & Lei, 1980). Furthermore, because adolescents face significant developmental challenges of autonomy and independence from parents and identity formation (Erikson, 1963; Erikson, 1968), the use of external, job- related, and internal stressor sub-scales is justified. In addition, the use of sub scales rather than a single laige scale removes a potential problem in the structural equation modeling process known as identification. Each stressor subscale was fonned by taking the sum of the standardized items, the overall stressor composite used in the regression analyses was fonned by taking the sum of all of the items making up the three subscales. These results, and those in the measurement invariance section below, confirm the hypothesized factor structure of the Stressors or Life Events construct. Hypothesis 1. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 89 Three psychological distress factors—anxiety, depression, and somatic symptoms—were identified from the exploratory factor analysis. These factors accounted for 55% of the total variance among the thirteen items. Only items with a strong single factor were included in the subscales, therefore one of the thirteen psychological distress items, “I feel nervous or anxious,” was dropped (loading = .31 on anxiety and .26 on somatic symptoms). The internal consistencies for each of the three psychological distress subscales are; (a) depression «=.82 and .87 (b) anxiety a =.76 and .80 (c) somatic symptoms a = .61 and .74 in 1990-91 and 1991-92 respectively. Scores for each of the coping and psychological distress subscales were obtained by standardizing and averaging their standardized item components. The 22 coping items were hypothesized to load on eight factors. When constrained to have eight factors, all but one loaded as expected, the “sleep more” item had two significant loadings. Because this item represents a behavior, it was placed with the behavioral factor. Antisocial coping, rather than with the Cognitive Restructuring factor. The eight coping factors accounted for 63% of the total item variance. The subscales and items are shown in Tkble 16 on page 157. Cronbach’s a ’s for these factors are given in the bottom group in Tkble 8 on page 90. Because coping was hypothesized as two types, approach and avoidance, an exploratory factor analysis was performed using the coping subscales. With Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 90 T able 8. Cronbach’s Alpha for M easured Variable Sub-Indices and Factor Indices Index # of Items 1990-91 M easured Variable Sub-Indices External Stressor 3 .46 Job Stressors 2 .45 Social Stressors 5 .51 Do Drugs Coping 3 .76 Social Coping 2 .73 Anti-social Coping 4 .66 Depression 5 .82 Anxiety 4 .76 Somatic Symptoms 3 .67 Exercise Coping 3 .77 Get Infonnation Coping 2 .62 Relaxation Coping 3 .63 Factor Indices Stressors Overall 10 .63 Avoidance Coping Overall 9 .77 Distress Overall 12 .87 Approach Coping Overall 7 .70 1991-92 .67 .64 .52 .70 .68 .65 .87 .80 .74 .80 .59 .61 .68 .73 .90 .73 Note: Values in the columns labeled 1990-91 and 1991-92 are Cronbach’s a ’s for the indices from the indicated year. The first group o f are for the analysis variables for the structural equation modeling procedure, the second group are for the higher order factors or constructs. two factors hypothesized a prioii, the exploratory factor analysis yielded a distinctive pattern of factor loadings, with most coping subscales having a high loading on only one factor. Based on this analysis, the cognitive restructuring and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 91 passive activities coping subscales were eliminated because they had relatively high loadings on both the adaptive and maladaptive coping factors. In summary, factor analysis yielded a total of 12 subscales for structural equation modeling (three stressor, six coping, and three psychological distress) created from thirty-eight questionnaire items. The subscales and items are shown in Tkble 16 on page 157. These results also provide support for Hypothesis 2, that two distinct coping factors exist, corresponding to Approach and Avoidance Coping (Billings & M oos, 1982; Holahan & M oos, 1991; M oos et al., 1990). For further support, see the measurement invariance section below. Three of the four second order constructs—approach coping, avoidance coping, and psychological distress—were created by taking the mean of the standardized questionnaire items. The stressors construct was created by summing the standardized item scores. These constructs are used for the regression analyses. It must be noted that these are not identical to the constructs in the structural equation models, which are implicitly created during the estimation and model specification process rather than having an explicit formula such as the sum of a group of standardized items. Cronbach’s a ’s for these factors are given in the bottom group in Table 8 on page 90. As expected because of the larger num ber of items making up each of these composites, a ’s are higher, with only Stressors having an a less than .7 (.63 and .68 in 1990-91 and 1991-92 respectively). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 92 Measurement Invariance Across Groups Confirmatory Factor Analysis A further step in assessing the adequacy of our hypothesized measurement model was performed to demonstrate that the variables we chose to represent the underlying latent constructs actually were statistically reliable reflections of the constructs. Cross-sectional confirmatory factor analysis (CFA) was used to assess the measurement model. In the initial CPA models the three latent constructs (Stressors, Avoidance Coping, and Psychological Distress) were hypothesized to generate the observed variation in the nine measured variables. Only the pure factor structure predicted a priori was allowed—each observed variable was allow ed to load on only one latent construct. The basic CFA models all had the following properties: (a) all factor variances were fixed at unity, (b) all hypothesized factor loadings were freed and all others were constrained to be zero, (c) all latent constructs were allowed to correlate freely. The CFA models were estimated using the entire sample, both Low and High Approach Coping groups. The basic models contained 24 degrees of freedom {dj). N either of the two initial models adequately fit the data, as both had X" values with p’s less than .001 (103.47 and 77.06 for 1990 and 1991 models respectively). However, both of the initial models captured a laige proportion of the sample variation, as shown by the values for the N Fl (.882 and .902 for 1990 and 1991 respectively) and the C Fl (.906 and .929 for 1990 and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 93 1991 respectively). As discussed previously, both N FI and CFI take on values between 0 and 1, with 0.9 considered an acceptable model fit for NFI. Model M odification To improve the model fit, selected free parameters were added to each of the initial models. Only certain classes of potential parameters were considered eligible to be added, with preference given to correlated errors within measures of the same latent construct, followed by correlated errors across constructs, followed by loadings of measured variables on factors other than those hypothesized initially. Only four additional parameters had to be added to each model, one additional factor loading and three correlated errors. This resulted in two final models (1990 and 1991), each with 20 degrees of freedom and values of 30.09 in 1990 (p=.068, NFI = .966, CFI = .988) and 28.47 in 1991 (p = .099, NFI = .964, C F I= .9 8 9 ). These final models adequately reflect the observed data, or at least may not be rejected as plausible explanations of the data. All hypothesized factor loadings were highly significant, confinning one major aspect of the hypothesized model. The basic factor structure from the CFA models is presented in Table 9 on page 94 and is built upon in the longitudinal single and multiple-group stmctural equation models. Because model modification is not a process specified a priori it may capitalize on chance to improve model fit. However, if the fundamental structure of the hypothesized model is not changed in the modified model, then the model Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 94 Table 9. Factor Structure of the Cross-Sectional CFA M easurement Model 1990-91 1991-92 Variable SFL RV SFL RV Stressors External Stressors .53 .85 .54 .84 Job Stressors .27 .96 .51 .86 Social Stressors .78 .47 .89 .51 Avoidance Coping Anti-Social Coping .90 .44 .82 .57 Do Drugs Coping .55 .84 .54 .84 Social Coping .24 .97 .27 .96 Psychological Distress Anxiety .77 .64 .77 .64 Depression .76 .66 .79 .61 Somatic Symptoms .72 .69 .71 .71 Note: Standardized solutions from the combined sample CFA models. SFL = Standardized Factor Loading. RV = Residual Variance. All factors were significant at / ; < .05. modification process merely confirms the structure represented in the initial, basic model. To test whether the basic structure of the model changed during the model-fitting process correlation coefficients were computed for the common set of 21 parameters between the initial and final models (nine factor loadings, nine error variances, and three factor inter-correlations). The correlation coefficient between the basic and final models in 1990 was .976, suggesting that the parameter estimates did not significantly change during the model modification process. In 1991 the results were slightly less convincing, with a correlation Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 95 coefficient of .817 between the param eter estimates in the basic and final models. However, these correlations were considered large enough to accept the measurement model as hypothesized. Regression M odels Three different sets of regression models were estimated, one set of models to test hypothesized mediator effects, and two sets to suggest or test hypothesized m oderator effects. They all can be considered subsets of the single Distress Outcome structural equation model. The first set of regression models (one for each Approach Coping group), presented in models numbered 2 and 3 in Tkble 10 on page 96, shows some support for the predicted mediation by Avoidance Coping of the Life Events to Distress relationship. There are three necessary conditions for a significant mediation effect to exist. First, the most distant antecedent variable must predict the mediator variable. In model 3, in both the Low and High Approach Coping groups. Time 0 Life Events significantly predicts Time 0 Avoidance Coping ((8 = .02, p < .05 for both models). Second, the m ediator variable must significantly predict the outcome variable. M odel 2 shows that Time 0 Avoidance Coping does significantly predict Time 1 Distress in both groups, with |S = .32 (p < .0001) in the Low Approach Coping group and /3 = .23 (p = .012) in the High Approach Coping group. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 96 T able 10. Regression M odels (Low versus High Approach Coping) Dependent Variable Independent Variables R Adj Life Life Avoidance EventSg Events^ Copingg Distressg Low Approach Coping iS (SE) 1 3 (SE) 0 (SE) 1 3 (SE) 1.Distress, -.OO(.Ol) .Ol(.Ol) ,07(.07) .63(.07)“ .62 .37 2 .Distress, .02(.01)+ NA .32(.07)" NA .34 .10 3 .Avoidance Copingg .02(.01)* NA NA N A .15 .02 High Approach Coping 1.D istress, -.O l(.O l) .04(.01)‘* -.04(.07) .59(.06)** .68 .45 2.D istress, .Ol(.Ol) NA .23(.09)* NA .20 .03 3 .Avoidance Copingg .02(.01)* NA NA N A .18 .03 Note; Values are regression coefficients with standard errors in parentheses. N A — Not applicable because the independent variable was not used in the regression. ^ — p<.lQ, * — p<.Q5, * * — p<.0001. Third, the mediated effect, given by the product of the two regression coefficients just mentioned, must be significant—using the z-ratio from equation (1) on page 36 to test for significance. In the Low Approach Coping group, the mediated effect is .0075 and the z-ratio is 1.91, so the effect is of marginal significance (p = .0 5 6 ). In the High Approach Coping Group, the mediated effect is .004, z= 1 .7 3 , .084, again of maiginal significance. Nonetheless, these Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 97 result suggest mediation of the Stressor-Distress relationship by Avoidance Coping. A direct effect of Stressors on Distress is of maiginal significance in the Low Approach Coping group and not significantly different from zero in the High Approach Coping group. The results for these tests of mediation also suggest a moderator effect, since the magnitude of the Avoidance Coping to Distress coefficient is smaller in the High Approach Coping group. Attenuation in one group and not the other suggests a moderator effect. This model, unlike the structural equation models, does not control for or model the effects of Time 0 Distress or Time 1 Life Events. The second set of regression models, numbered 1 in Ikble 10 on page 96, suggests that a moderator effect may exist. In these two regression models, conducted separately for the Low and High Approach Coping groups. Time 1 Distress is predicted by Time 0 and Time 1 Life Events, Time 0 Avoidance Coping, and T in e 0 Distress. As can be seen in Tkble 10, the only consistently significant predictor of Distress at Time 1 is Distress at Time 0. In the High Approach Coping group. Life Events at T m e 0 is also a statistically significant predictor, hence there may be a difference between the two groups. This finding suggests of a moderator effect, although theory predicts that the magnitude of the param eter estimate should be lower in the High Approach Coping group (a buffer effect), while the results show it to be greater in this sample (.04 versus .01). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 98 Also suggesting a moderator effect is the finding that the regression coefficient between Avoidance Coping at Time 0 and Distress at Time 1 is negative in the High Approach Coping group, but positive in the Low Approach Coping group. Neither of these parameters is statistically significant, but the difference could be significant—this cannot be tested in this model. T able 11. Tim e 1 Distress Outcome Regression Model with Interaction Terms Independent Variable 3 (Standard Error) Main Effects Terms Life EventSg -.00 (.01) Life Events, .02 (.01)** Avoidance Copingg -.02 (.05) Approach Copingg -.06 (.09) Distressg .63 (.05)** Interaction Terms Approach Copingg X LEg -.03 (.02) Approach Copingg X LE; .04 (.02) Approach Copingg X ACg -.12 (.05)* Approach Copingg X Distressg .07 (.08) R- .43 Interaction Terms Semipartial R ~ .014 Effect S ize" /- .024 Note: “Effect size is defined in the Statistical Power section on page 64. Values are regression coefficients with standard errors in parentheses. Dependent variable is Distress at Time 1. LE= Life Events, A C =A voidance Coping. = p < . 10, * = p < .05, * * = p < .0001. The third set of regression models augment the second set and constitute a stronger test of hypothesized moderator effects. This model, presented in Table 11 on page 98, adds Approach Coping cross-product tenns to the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 99 regression equation. These terms represent interaction effects and allow statistical testing for a moderator effect of Approach Coping on the Stressor-Avoidance Coping-Distress relationships. This model is tested on the full analysis sample of 401. The outcome variable is Distress at Time I. The main independent variables included in this model are lim e 0 and Time 1 Life Events, T m e 0 Approach Coping, Time 0 Avoidance Coping, and T m e 0 Distress. The interaction tenns are fonned by the cross-product of the putative moderator variable, T m e 0 Approach Coping, and each of the other predictor variables. Table 11 on page 98 shows the results for this model. The only main effects tenn with significant regression coefficients are Life Events at T m e 1 and baseline Distress. Because Avoidance Coping at Time 0 is not a significant predictor of Distress at Time 1 mediation by Avoidance Coping of the Stressor- Distress effect is not supported. However, these results are supportive of a stress- buffer effect for Approach Coping. A significant negative coefficient for the interaction term Approach Coping X Avoidance Coping indicates that at higher levels of Approach Coping, the adverse effect of Avoidance Coping is attenuated. The lack of a significant direct effect for Avoidance Coping at Time 0 may be because it only has a weak direct effect or because it affects Distress more proximally, i.e. Avoidance Coping may exert its effect on Distress at Time 1 through Distress at Time 0. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 100 Structural Equation Models Introduction Once the measurement model was specified, we tested the across-time causal relations between the latent variables. Structural or path models were created to test these relationships. These models built on the CFA or m easurement model presented above but included regression coefficients representing unidirectional influences of one factor on another. Other, non-causal associations were modeled as correlations between factors. O f most interest are regression effects that allow us to make plausible causal inferences. Because some o f the correlated errors or factor loadings added during the model modification process to derive the final CFA model may actually represent across-time regression effects the structural models all built upon the basic measurement model represented in the initial CFA, without the added correlated errors or factor loadings. To test the m ajor hypotheses concerning group differences between those adolescents with low versus high approach coping skills, the models were specified as two group models. The groups were determined using a median split on the Approach Coping composite in 1990-91 which resulted in a sample of 195 in the low approach coping group and 206 in the high approach coping group. Several models were examined, but the main model of interest contains the following latent constructs: Stressors in both 1990-91 and 1991-92, Avoidance Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 101 Coping in 1990-91, and Distress in both 1990-91 and 1991-92. This model is tanned the “Distress Outcome M odel” to distinguish it from the others. The Fully Lagged Model adds Avoidance Coping in 1991-92 to the Distress Outcome M odel. The Change Scores M odel uses change or difference scores for Avoidance Coping and Distress instead of including both baseline and follow-up measures in the model. Additionally, to allow comparison with previous work (Broder et al., under review), two cross-sectional models were investigated, one in 1990-91 and one in 1991-92. All five final SEM models are summarized in Table 14 on page 132. Distress Outcome Model Important aspects of the Distress Outcome model as presented in Figure 5 on page 158 (and in standardized form in Figure 6 on page 159) include the structural parameters estimated. In particular, Time 0 Life Events was modeled as causally influencing Time 0 Avoidance Coping and Distress at both Time 0 and Time 1. In addition. Distress at Time 1 had antecedents of all of the other constructs: Life Events, Avoidance Coping, and Distress at Time 0 and Life Events at Time 1. Finally, because some individuals may be better at remembering or reporting stressful life events we hypothesized that Life Events at Time 0 and T m e 1 are correlated. However, this association was not modeled as a causal one because the assumption that the occurrence of a particular life event in 1990 will cause the occurrence o f other stressful life events in 1991 seems too Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 102 rigorous (although counter-examples can be found, e.g. perhaps moving to a new home in 1990 could cause more aiguments with one’s family in 1991). Also, Time 0 Avoidance Coping was allowed to correlate with Time 0 Distress. Separate M odels by Group The initial step in the analysis was to estimate the basic SEM model separately for the two groups: low and high avoidance coping, with sample sizes of 195 and 206, respectively. As seen in Thble 12 on page 103, the initial longitudinal SEM models for the two groups (M^g and M ^q) have inadequate fit, X"(82) = 194.00 and 269.75 (p < .0001), for the High and Low Approach Coping groups respectively. Even the fit indices for these models show an inadequate fit, with the N FI in the .7 to .8 range and the CFI only slightly higher (values greater than .9 empirically demonstrate an adequate fit). Because the p-values of the x" test statistic for both of these models is less than .05, we reject the null hypothesis that the relationships represented by these data do not significantly differ from those predicted by the model. That is, these models did not adequately represent the observed covariation in these data. Model Modification To improve the fit of these models, certain residual variances or errors were allowed to correlate. To ensure that more theoretically defensible models were derived, no other model modifications such as additional factor loadings or correlated disturbances were permitted. Although the LM test suggested Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 103 T able 12. M ultiple Group Modal Development (Low versus High Approach Coping) Longitudinal SEM Model - Distress Outcome c n Groun Y " df p NFI H ieh Approach Copine ^HO 194.00 82 < .001 .802 .872 M hi 97.29 76 .050 .901 .976 Low Approach Copine ^LO 269.75 82 <.001 .727 .788 M li 90.22 73 .084 .909 .981 Combined Two Grouo Model MO 187.50 149 .018 .905 .978 M l 211.78 159 .003 .893 .970 M l-M O 24.28 10 .007 M l* 199.78 158 .014 .899 .976 M1*-M0 12.28 9 .198 M2 209.60 168 .016 .894 .976 M2-M1* 9.83 10 .456 M3 229.87 176 .004 .883 .969 M3-M2 20.27 8 .009 M3* 219.30 174 .011 .889 .974 M3*-M2 9.70 6 .138 M ^o M[^q = basic theoretical models for high and low adaptive coping groups, respectively. M y; and M ^] = modified models with correlated errors suggested by the LM test on M^^q and M ^q, respectively. MO = basic model combining high and low adaptive coping groups. M l = model with all means (including factor means) constrained equal across groups. M2 = model with all factor loadings constrained equal across groups over MP. M3 = model with all factor variances and covariances and/or regression weights constrained equal across groups over M2. * = model with relevant parameters partially constrained equal across groups. Further explanation o f the model Fitting process is given in the text.__________ Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 104 correlated errors to add, we were strict in trying to only add those errors that were specified in advance as being substantively meaningful. For example, the "best” type of correlated error to add was correlations between the same measured variable at different times, e.g. Time 0 Job Stressors and Time 1 Job Stressors. As shown in Ikble 13 on page 105, many of the correlated errors added represent method or response effects between either the same variable measured at different times or different variables measured in similar formats. In fact, it would have been a reasonable model development alternative to hypothesize a priori that correlated errors existed between all variables measured at two different times. In the Low Avoidance Coping group model four of the nine correlated errors added were between the same variables measured at different times and the other five were between different variables representing the same factor. A sim ilar pattern is evident in the High Avoidance Coping group, except that a single correlated error, between Social Stressors in 1990 and Anxiety in 1990 is not of either of these two types. Correlation coefficients between the two sets of factor loadings, regression weights, and inter-factor correlations in the initial and final Distress Outcome models were .985 and .989 in the High and Low Approach Coping Groups respectively, indicating that the model modification process did not substantially change the substantive parameters of the models. Both of the modified single Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Table 13. Correlated Errors in the Final Single Group "Distress Outcom e” 105 Models Low Avoidance Coping High Avoidance Coping Across Time. Same Variables I r Job Stressors .34 Job Stressors .20 Depression .46 Depression .41 Somatic Symptoms .43 Somatic Symptoms .43 External Stressors .27 Anxiety .32 W ithin Factors. Different Variables ExtStressg with Job Stressq .25 ExtStresSj with Job Stress^ .22 ExtStressg with Social Stressg .62 DoDrugSp with Social Copingp .28 Job Stressg with Social Stressg .22 DoDrugSg with Antisocial Copingg .28 Across Factors Social StressQ with Anxietyg .35" Note; Values are the correlations between residual variances, "correlated erro rs.” All were significant at p < .05. "This is the only correlated error that is not within a single factor (either at one time or between tim e 0 and tim e 1). group models, M ^i and M[^,, were statistically acceptable and the CFI and NFI indices showed a good fit to the observed data. A summary of the model development process for the Distress Outcome model, including the initial and final single group models, is shown in Thble 12 on page 103. The two final single group models, M ^i and Mp^,, are combined to form the initial multiple- Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 0 6 group model, MO, whose degrees of freedom and x ' values can be calculated by summing those from the final single group models. Equality of Variable Means Across Groups As the mean levels of the factors are more interesting than the mean levels of the observed variables, in the first multiple-group model equality constraints are imposed on the variable means, but the factor means are allowed to be free. This results in (p - f) additional degrees of freedom, where p is the number of variables (15 for this model) and f is the number of factors (5). M odel M l therefore has 10 more degrees of freedom than Model MO. Because M odel M l is nested within MO, the constraints added can be evaluated by taking a difference of the X " values of M l and MO. The result is distributed as a x^ statistic with degrees of freedom equal to the difference in the degrees of freedom between the two models. It represents the null hypothesis that all added constraints are reasonable. As seen in Ikble 12 on page 103, overall equality of variable means is an unreasonable constraint (Ml-MO x^(10)=24.28, p —.QQl). The LM test indicated that the mean of Social Coping in 1990 could not be assumed to be equal across the two groups. Model M l* releases this constraint and yields a goodness-of-fit x" value of 199.78 on 158 degrees of freedom, which is more comparable to the basic model MO than was M l. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 107 Equality of Factor Loadings Across Groups M odel M 2, developed from and nested within M odel M l*, is used to test whether the factor loadings are equal across the two groups. There are 15 factor loadings in the SEM model, but five of them, one per factor, are fixed at unity for model identification purposes. M odel M 2, with across-group equality constraints on factor loadings, therefore has 10 additional degrees of freedom over Model M l*. In this case, based on the difference test between M2 and M l*, we cannot reject the null hypothesis and thus accept all the constraints on the factor loadings. That is, the hypothesis that the factor structure represented by the factor loadings is the same in the two approach coping groups adequately represents these data. Equality of Factor Correlations and Regression Weights In the multiple group models previously discussed we have shown partial invariance (equality) of the measurement model across the two Approach Coping groups. Relative invariance of the measurement model is a precondition for testing the substantively interesting hypotheses. These include hypothesized inequality of regression coefficients between the two groups. M odel M3 imposed a total of eight additional equality constraints on M odel M 2. Two of these represented constraints on the correlations between Life Events at Time 0 and Time 1 and between Avoidance Coping and Distress at Time 0. The other six represent equality constraints on the regression coefficients in the model. The Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 108 difference test between M odels M 3 and M 2, x'(8)= 2 0.27, p = .009, indicates that not all of the equality constraints are appropriate. The LM test suggested removal of the equality constraint on the regression coefficient between Avoidance Coping at Time 0 and Distress at Time 0. A fter this constraint was removed the modified M odel M3* still contained some unreasonable constraints. The LM test suggested removing the constraint on the mean of the measured variable Social Life Events at Time 1. Although this was not a constraint added in the most recent model m odification it was freed, yielding M odel M3*. All rem aining constraints—representing partial invariance of both measurement and structural relationships across groups—were acceptable, with a % - difference test of X‘(6 )= 9 .7 0 , /i = .138. The final model M3* did not provide an acceptable fit to the data, x '(1 7 4 )= 2 1 9 .3 0 , /?= .011. However, the N FI and CFI indices show that the model represents a large portion of the variation in these data. Summary of Distress Outcome Longitudinal SEM M odel In summary, analysis of the Distress Outcome multiple group structural equation model showed support for several differences in both the measurement and structural models between the Low and High Approach Coping groups. O f the set of correlated errors added to the two separate single group models before combining them to form the multiple group model, four correlated errors were added to both the Low and High Approach Coping groups, two to only the High Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 109 Approach Coping group, and five to only the Low Approach Coping group out of a total of 105 possible correlated errors. The only significant differences between groups in the measurement model were mean differences of the measured variables Social Coping at Time 0 and Social Stress at Time 1, with the High Avoidance Coping group having higher mean values for both of these variables. In addition, only one factor level mean difference existed, providing evidence for Hypothesis 7. The mean levels of the Avoidance Coping at Time 0 construct was greater in the High Approach Coping group. The Distress Outcome structural model is summarized in Ihble 14 on page 132 and in Figure 5 on page 158. In both groups the regression coefficient between Life Events at Time 0 and Distress at Time 0 is significant and positive. Also significant in both groups and positive are the regression coefficients between Life Events at Time 0 and Avoidance Coping at T m e 0, between Life Events at Time 1 and Distress at Time 1, and between Distress at Time 0 and Distress at Time 1. This means that the Life Events constructs, measuring major life events occurring in the previous 12 months, significantly predict psychological distress at the end of the 12 months. However, because there is no significant relationship between Life Events at Time 0 and Distress at Time 1, there is no direct effect of m ajor life events occurring 12 to 24 months ago predicting current psychological distress. As expected, increased levels of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 110 Psychological Distress at Time 0 predicted increased Distress at Time 1. Furthermore, also as expected. Life Events at Time 0 were significantly correlated with Life Events at Time 1. These significant parameters provide support for Hypothesis 3, that the Life Events to Distress regression coefficient is greater than 0 (although this is only true for Life Events in the immediately prior 12 months) and partial support for Hypothesis 5, i.e. that the first leg of the mediation effect is true. Some support exists for the hypothesized mediation of the effect of life events on psychological distress, since in both Low and High Approach Coping groups the indirect effect of Life Events at T in e 0 on Distress at Time 1 was significant. However, since the regression coefficient between Avoidance Coping at Time 0 and Distress at T m e 1 is not significantly different from 0 in the Low Approach Coping group, the major mediation seems to be occurring by way of Life Events at Time O’s effect on Distress at T m e 0 or its correlation with Life Events at Time 1 rather than through an effect on Avoidance Coping. Hypothesis 4 is not proved, for in neither of the two groups is the regression coefficient between Avoidance Coping at T m e 0 and Distress at Time 1 significantly greater than 0. The regression coefficient between Avoidance Coping and Distress is lower in the High Approach Coping group than in the Low Approach coping group. That is, the adverse effects of Avoidance Coping are attenuated in the High Approach Coping group. This is one of the possible formulations of the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. Ill classic buffer effect. However, the mean level of psychological distress is not significantly different between the two groups. As will be discussed later, this may be due to other, unmeasured, group differences in areas such as self- awareness or abstract cognition. Therefore, although increased levels of Approach Coping do seem to provide protection against the adverse effects of either increased levels of Avoidance Coping or increased levels of Life Events, Hypothesis 6 is only partially proved. Fully Lagged Model All of the m ajor parameters from the Distress Outcome model presented on page 101 are included in the Fully Lagged model shown in Figure 7 on page 160 and in Figure 8 on page 161. In addition, Avoidance Coping at Time 1 was hypothesized to correlate with Distress at Time 1 and to have causal antecedents of Life Events at Times 0 and 1 and Avoidance Coping at Time 0. This model, and succeeding ones, are presented in less detail than the prim ary Distress Outcome model. Separate Models by Group The initial step in the analysis was to estimate the basic SEM model separately for the two groups: low and high avoidance coping, with sample sizes of 195 and 206, respectively. In the Fully Lagged model, the initial longitudinal SEM models for the two groups (M^o and M ^q) had inadequate fit, X‘(122)=309.92 and 446.97 ( p < .001), for the High and Low Approach Coping Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 112 groups respectively. Even the fit indices for these models show an inadequate fit, with the N FI in the .6 to .7 range and the CFI only slightly higher. These models do not adequately represent the observed covariation in these data. Model M odification As discussed previously, to improve the fit of these models, certain residual variances or errors were allowed to correlate. Of a possible 153 correlated errors, 14 were added to the High Approach Coping group and 19 to the Low Approach Coping group. Three of the correlated errors in the High Approach Coping group and two in the Low Approach Coping group represent neither same variable/different time nor within construct correlated errors. All five of these were between avoidance coping and distress variables within a single wave and so are preferable to across wave correlated errors between antecedent and consequent constructs. Correlation coefficients between the two sets of factor loadings, regression weights, and inter-factor correlations in the initial and final Fully Lagged models were .967 and .973 in the High and Low Approach Coping Groups respectively, indicating that the model modification process did not substantially change the substantive parameters of the models. Both of the modified single group models. M u and M ^ , were statistically acceptable and the CFI and NFI indices showed a good fit to the observed data. For M odel M ^ ,, x^( 108) = 131.68, p = .060, CFI = .895, and NFI = .978. For Model M ^ , x'(103) = 126.86, p=.055, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 113 C F I= .9 0 0 , and N F I= .9 7 9 . The two final single group models, M ^i and M ^ ,, are combined to form the initial multiple-group model, MO, whose degrees of freedom and values can be calculated by summing those from the final single group models. Equality of Variable M eans Across Groups In the Fully Lagged model, after equality constraints across groups are imposed on measured variables. M odel M l has 12 more degrees of freedom than M odel MO. Because M odel M l is nested within MO, the constraints added can be evaluated by taking a difference of the values of M l and MO. For the Fully Lagged M odel, overall equality of variable means is an unreasonable constraint (M l-M O x '(1 2 )= 2 5 .7 9 , p < .02). The LM test indicated that the mean of Drug Use Coping in 1990 could not be assumed to be equal across the two groups. Model M l* releases this constraint and yields a goodness-of-fit x" value of 273.92 on 222 degrees of freedom , which is more comparable to the basic model MO than was M l—the x' difference test is non-significant, x^(H ) = 15.80, p > .1. Equality of Factor Loadings Across Groups M odel M 2, developed from and nested within Model M l’, is used to test whether the factor loadings are equal across the two groups. There are 18 factor loadings in the SEM model, but six of them, one per factor, are fixed at unity for model identification purposes. M odel M2, with across-group equality constraints on factor loadings, therefore has 12 additional degrees of freedom over M odel Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 114 M l*. In this case, based on the difference test between M2 and M l*, we cannot reject the null hypothesis and thus accept all the constraints on the factor loadings. That is, the hypothesis that the factor structure represented by the factor loadings is the same in the two approach coping groups adequately represents the data used in the Fully Lagged model. Equality of Factor Correlations and Regression Weights In the multiple group models previously discussed we have shown partial invariance (equality) of the measurement model across the two Approach Coping groups. Relative invariance of the measurement model is a precondition for testing the substantively interesting hypotheses. These include hypothesized inequality of regression coefficients between the two groups. M odel M3 imposed a total of 13 additional equality constraints on Model M 2. Three of these represented constraints on the correlations between Life Events at Time 0 and Time 1 and between Avoidance Coping at Time 0 and Distress at Time 0 and between Avoidance Coping at Time 1 and Distress at Time 1. The other 10 represent equality constraints on the regression coefficients in the model. The x‘ difference test between M odels M3 and M 2, x^(13)=34.42, p < .01, indicates that not all of the equality constraints are appropriate. That is, because the difference test is statistically significant we must reject the null hypothesis that all of the added constraints are consistent with these data. Guided by the LM test we removed the equality constraints on the regression coefficient between Avoidance Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 115 Coping at Tune 0 and Distress at Time 0 and on the correlation coefficient between Avoidance Coping at Time 1 and Distress at Time 1. A fter these constraints were removed the modified M odel M3* still contained some unreasonable constraints. The LM test suggested removing the constraint on the mean of measured variable Antisocial Coping at Time 0. Although this was not a constraint added in the most recent model modification it was freed, yielding Model M3*. All remaining constraints—representing partial invariance of both measurement and structural relationships across groups—were acceptable, with a difference test of x^(10) = 15.59, p > A . The final model M3* did not provide an acceptable fit to the data, x '(2 4 4 )=309.97, p = .0 0 3 . However, the NFI and CFI indices, as shown in Thble 14 on page 132, show that the model represents a laige portion of the variation in these data. Summary of Fullv Lagged Longitudinal SEM Model In summary, analysis of the Fully Lagged multiple group structural equation model showed support for several differences in both the measurement and structural models between the Low and High Approach Coping groups. Of the set of correlated errors added to the two separate single group models before combining them to form the multiple group model, ten correlated errors were added to both the Low and High Approach Coping groups, four to only the High Approach Coping group, and nine to only the Low Approach Coping group out of a total of 153 possible correlated errors. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 1 6 The only significant differences between groups in the measurement model were mean differences of the measured variables Antisocial Coping at Time 0 and Do Drugs Coping at Time 0, with the High Avoidance Coping group having higher mean values for both of these variables. Only one factor level mean difference existed. The mean levels of the Avoidance Coping at Time 0 construct was greater in the High Approach Coping group; however, there were no significant differences on the means of this construct at Time 1. The Fully Lagged structural model is summarized in Tkble 14 on page 132 and in Figure 5 on page 158. In general, most of the regression coefficients that were significant in the Distress Outcome model were significant in the Fully Lagged model. A notable exception is that in neither the Low nor High Approach Coping group did Avoidance Coping at Time 0 significantly predict Distress at Time 1. Both Life Events at T m e 1 and Avoidance Coping at Time 1 did significantly predict Avoidance Coping at Time 1. Some support exists for the hypothesized mediation of the effect of life events on psychological distress, since in both Low and High Approach Coping groups the indirect effect of Life Events at Time 0 on Distress at Time 1 and on Avoidance Coping at T m e 1 were both significant. However, since the regression coefficient between Avoidance Coping at T m e 0 and Distress at Time 1 is not significantly different from 0 in either group, the major mediation seems to be occurring by way of Life Events at T m e O’s effect on Distress at T m e 0 or Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 117 its correlation with Life Events at Time 1 rather than through an effect on Avoidance Coping. Hypothesis 4 is not proved, for in neither of the two groups is the regression coefficient between Avoidance Coping at Time 0 and Distress at Time 1 significantly greater than 0. Avoidance Coping at Time 1 and Distress at Time 1 were significantly correlated only in the High Approach Coping group. The Fully Lagged model does not provide good support for the moderator o r buffer effect. However, as mentioned above, these types of models may be difficult to interpret empirically as the residual variance left to be explained by the portions of the model that are of greatest interest may be small due to the large amount of variance explained by the prediction Time 1 constructs by Time 0 constructs. Change Scores Model The Change Scores Model is shown in Figure 9 on page 162. The parameters in this model include Life Events at Time 0 and T m e 1, an Avoidance Coping change score, and a Distress change score. Change scores are the difference between 1991-92 values and 1990-91 values. The two Life Events constructs are hypothesized to be correlated and both are hypothesized to predict the two change score constructs. A regression coefficient between change in Avoidance Coping and change in Distress is also modeled. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 118 Separate Models by Group In the Change Scores model, the initial longitudinal SEM models for the two Avoidance Coping groups separately (M^o and M lq) had inadequate fit, X"(48)=77.00 (p = .005) and % ^(48) = 125.96 (p < .001), for the High and Low Approach Coping groups respectively. The fit indices for these models are also inadequate, with the NFI in the .7 range and the CFI near .85. These models do not adequately represent the observed covariation in these data. M odel Modification To improve the fit of these models, certain residual variances or errors were allowed to correlate. O f a possible 66 correlated errors, 2 were added to the High Approach Coping group and 4 to the Low Approach Coping group. One of the correlated errors in the High Approach Coping group is between the same Life Events variable at Time 0 and Tune 1, the other is between two variables representing the same construct at the same wave. Three of the correlated errors in the Low Approach Coping group are same variable/different time errors, and the fourth is between two variables measuring the same construct, but at two different times. Correlation coefficients between the two sets of factor loadings, regression weights, and inter-factor correlations in the initial and final Change Scores models were .997 in the High Approach Coping Group, but only .666 in the Low Approach Coping Group. However, most of the difference between initial and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 119 final sets of parameters in the Low Approach Coping group was because of a laige difference in the correlation between the Life Events constructs at Tunes 0 and 1. Without this single parameter, the correlation between the remaining parameters in the initial and final models was .859, indicating that the model modification process did not substantially change the other substantive parameters of the models. Both of the modified single group models, and Mpj,, were statistically acceptable and the CFI and N FI indices showed a good fit to the observed data. For Model M^^,, x '(4 6 )= 5 6 .0 0 , /? = .148, C F I-.8 2 5 , and NFI = .961. For Model M ^i, x^(44)=58.84, /?= .067, C F I-.8 5 7 , and N F I-.9 5 7 . The two final single group models, M ^ and Mp^,, are combined to fonn the initial multiple-group model, MO, whose degrees of freedom and x" values can be calculated by summing those from the final single group models. Equality of Variable Means Across Groups In the Change Scores model, after equality constraints across groups are imposed on measured variables. Model M l has 8 more degrees of freedom than Model MO. In this case, overall variable mean invariance is a reasonable constraint (Ml-MO x'(8) = 10.99, p > .2). Equalitv of Factor Loadings Across Groups Model M 2, developed from and nested within M odel M T , is used to test whether the factor loadings are equal across the two groups. There are 12 factor loadings in the SEM model, but four of them, one per factor, are fixed at unity Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 2 0 for model identification purposes. M odel M 2, with across-group equality constraints on factor loadings, therefore has 8 additional degrees of freedom over M odel M l*. Overall invariance of factor loadings across groups was an unreasonable constraint, with %-(8)=23.23, p<.Q\. After releasing the equality constraint on the factor loading of Life Events at Time 0 on External Stressors at Time 0, some of the remaining constraints were still unsupported by the data. By releasing the constraint on equality of the mean of Social Stressors at lim e 1 across groups, an acceptable model M 2’ was found, with M 2’-M l* having x ‘(6) = 10.13, p > .\. Equality of Factor Correlations and Regression Weights In the multiple group models previously discussed we have shown partial invariance (equality) of the measurement model across the two Approach Coping groups. M odel M3 imposed a total of 6 additional equality constraints on Model M2*. One of these represented a constraint on the correlation between Life Events at Time 0 and at T m e 1. The other 5 represent equality constraints on the regression coefficients in the model. The difference test between Models M3 and M2*, x^(6) = 10.04, p > .\ , indicates that all of the equality constraints are reasonable. That is, there were no structural differences between the Low and High Approach Coping groups in the Change Scores model. The final model M3 did not provide an acceptable fit to the data, x “(l 10) = 146.00, p — .Q\2. However, Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 121 the NFI and CFI indices, as shown in Thble 14 on page 132, show that the model represents a laige portion of the variation in these data. Summary of Change Scores Longitudinal SEM M odel In summary, analysis of the Change Scores multiple group structural equation model only showed support for differences in the measurement model between the Low and High Approach Coping groups. O f the set of correlated errors added to the two separate single group models before combining them to form the multiple group model, one correlated error was added to both the Low and High Approach Coping groups, one to only the High Approach Coping group, and three to only the Low Approach Coping group out of a total of 66 possible correlated errors. The only significant differences between groups in the measurement model was a higher level of Social Life Events at Time 1 in the High Avoidance Coping group and a lower factor loading of Life Events at Tune 0 on External Life Events at Time 0 in the High Avoidance Coping group. No significant factor level mean differences existed. The mean level of the Avoidance Coping change score was lower in the High Avoidance Coping group, although this was of only maiginal significance. This is consistent with the Fully Lagged model results, which found no difference between groups in Avoidance Coping levels at Time 1, but higher levels in the High Avoidance Coping group at Time 0. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 122 The Change Scores structural model is summarized in Tkble 14 on page 132 and in Figure 9 on page 162. In general, this model was difficult to interpret, with a negative regression coefficient between Life Events at Time 0 and change in Distress. This is likely because high levels of Life Events at Time 0 are associated with higher levels of Distress at Time 0, so a ceiling effect combined with regression to the mean could be taking place with lower levels of change in Distress because some students cannot report Distress any higher at Time 1 than they have at Time 0. No significant indirect effects were found in the Change Scores model. Furtherm ore, because there is no regression path between Life Events at Time 0 and the Avoidance Coping change score, the only possible indirect or mediated path would be from Life Events at Time 0 to change in Distress, mediated by Life Events at Time 1, which is not a very substantively interesting path, for it provides little more guidance in designing prevention programs than the mere existence of the regression path between Life Events at Time 1 and change in Distress. The Change Scores model does not provide good support for the moderator or buffer effect, since no structural differences exist between the Approach Coping groups. One problem with this model is related to the use of a change score for Avoidance Coping. For some students, change in avoidance coping does not accurately reflect their use of these maladaptive coping strategies. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 123 since individuals who consistently make greater or lesser use of these strategies would both have low change scores, regardless of whether they used or did not use these strategies. Another problem with change score models is the implicit assumption that changing from using zero Avoidance Coping strategies to using one is the same as changing from using five strategies to using six. Finally, and most importantly, the Change Scores model is less useful because of its relative lack of utility in providing insight into the development of prevention programs. It is more difficult to imagine designing programs whose goal is to change students change scores than it is to change the underlying constructs. 1990 Cross-Sectional Model Two cross-sectional models were evaluated in this study, one using 1990 data shown in Figure 10 on page 163 and one using 1991 data shown in Figure 11 on page 164. The 1990 model is presented first. Separate M odels by Group In the 1990 Cross-Sectional model, the initial longitudinal SEM models for the two Avoidance Coping groups separately and M lq) had inadequate fit, x '(2 4 )= 70.05 and 73.42 ( p < .001), for the High and Low Approach Coping groups respectively. The goodness-of-fit indices for these models were somewhat better than what the with the NFI equal to .85 and .84 and the CFI equal to .89 and .88 for the High and Low Approach Coping groups respectively. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 124 Nonetheless, these models do not adequately represent the observed covariation in these data. Model Modification O f a possible 36 correlated errors, four were added to each of the High and Low Approach Coping groups. One of the correlated errors in each group was between an Avoidance Coping measured variable and a Distress measured variable. The other three correlated errors were all between indicators of the same construct. Correlation coefficients between the two sets of factor loadings, regression weights, inter-factor correlations, factor variances, and error variances in the initial and final 1990 Cross-Sectional models were .998 and .994 in the High and Low Approach Coping groups, respectively, indicating that the model modification process did not substantially change the substantive parameters of the models. Both of the modified single group models, M ^j and Mp^,, were statistically acceptable and the CFI and N FI indices showed a good fit to the observed data. For Model M j^,, %^(20)-2 8 .7 8 , p = .092, CFI = .938, and N F I= .980. For Model M ^ , % % = 2 6 .0 7 , p = .1 6 3 , CFI = .942, and NFI = .985. The two final single group models, M ^ and Mj.j,, are combined to form the initial multiple-group model, MO. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 125 Equalitv of Variable Means Across Groups In the 1990 Cross-Sectional model, after equality constraints across groups are imposed on measured variables, M odel M l has 6 more degrees of freedom than M odel MO. For the 1990 Cross-Sectional model, observed variable mean invariance is an unreasonable constraint (M l-M O %^(6) = 16.92, /?< .0 1 ). The LM test indicated that the mean of Social Coping in 1990 could not be assumed to be equal across the two groups. Model M l* releases this constraint and yields a goodness-of-fit x^(45)=59.88, which is more comparable to the basic model MO than was M l—the x ' difference test is non-significant x^(5)= 5.03, p > .3. Equalitv of Factor Loadings Across Groups Model M 2, developed from and nested within M odel M l*, is used to test whether the factor loadings are equal across the two groups. There are 9 factor loadings in the SEM model, but three of them, one per factor, are fixed at 1 for model identification purposes. Model M 2, with across-group equality constraints on factor loadings, therefore has 6 additional degrees of freedom over Model M l*. In this case, based on the difference test between M 2 and M l*, x^(6) = .825, p > .99, we accept all the constraints on the factor loadings. Equality of Factor Correlations and Regression Weights In the multiple group models previously discussed we have shown partial invariance (equality) of the measurement model across the two Approach Coping groups. M odel M3 imposed a total of 3 additional equality constraints on Model Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 126 M 2. These constrain the regression coefficients in the model equal across Approach Coping groups. The difference test between M odels M3 and M2, X^(3)— 5.99, p > .\ , indicates that all of the equality constraints are reasonable. That is, there were no structural differences between the Low and High Approach Coping groups in the 1990 Cross-Sectional model. The final model M 3 fit the data adequately, x '(5 4 )= 6 6 .6 9 , p = .l2. The N FI and CFI indices, as shown in Tkble 14 on page 132, also show that the model represents a laige portion of the variation in these data. Summary of 1990 Cross-Sectional Longitudinal SEM Model In summary, analysis of the 1990 Cross-Sectional multiple group structural equation model only showed support for differences in the measurement model between the Low and High Approach Coping groups. O f the set of correlated errors added to the two separate single group models before combining them to fonn the multiple group model, two correlated errors were added to both the Low and High Approach Coping groups, and two each to the High and Low Approach Coping groups separately, and three to only the Low Approach Coping group out of a total of 36 possible correlated errors. The only significant difference between groups in the measurement model was a higher level of Social Coping at Time 0 in the High Avoidance Coping group. As in other models, the mean level of Avoidance Coping was higher in the High Avoidance Coping group. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 127 The 1990 Cross-Sectional structural model is summarized in Table 14 on page 132 and in Figure 10 on page 163. This model, which did not control for preexisting differences in the Distress outcome variable, showed support for direct positive regression coefficients between Life Events and Avoidance Coping, Life Events and Distress, and Avoidance Coping and Distress. The indirect or mediated effect of Life Events on Distress was significant in the 1990 Cross-Sectional model. As there is only one indirect path in the model, the indirect effect necessarily occurs through mediation by Avoidance Coping. That is, some of the increase in Distress that occurs because of increases in Life Events occurs because increased experience of stressful life events causes increased used of Avoidance Coping, which in turn causes increased psychological distress. The 1990 Cross-Sectional model does not provide good support for the moderator or buffer effect, since no structural differences exist between the Approach Coping groups. 1991 C ross-Sectional M odel Separate M odels by Group In the 1991 Cross-Sectional model, the initial longitudinal SEM models for the two Avoidance Coping groups separately (M^o and M^g) had inadequate fit, x^(24)=62.99 and 52.38 (p < .001), for the High and Low Approach Coping groups respectively. The goodness-of-fit indices for these models were somewhat Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 128 better than suggested by the values, with the NFI equal to .86 and .89 in the High and Low Approach coping groups respectively and the CFI equal to .93 in both groups. Nonetheless, these models do not adequately represent the observed covariation in these data. Model Modification O f a possible 36 correlated errors, four were added to the High Approach Coping group and two to the Low Approach Coping group. Two of the four correlated errors in the Low Approach Coping group were between a Life Events measured variable and an Avoidance Coping measured variable. The other two correlated errors were all between indicators of the same construct. In the High Approach Coping group, one correlated error was between indicators of the same construct and the other was between an indicator of the Avoidance Coping construct and an indicator of the Distress construct. Correlation coefficients between the two sets of factor loadings, regression weights, inter-factor correlations, factor variances, and error variances in the initial and final 1991 Cross-Sectional models were 1.000 and .994 in the High and Low Approach Coping groups, respectively, indicating that the model modification process did not substantially change the substantive parameters of the models. Both of the modified single group models, M ^ and Mf^,, were statistically acceptable and the CFI and NFI indices showed a good fit to the observed data. For Model M ^ i, x^(20)=26.978, p = .1 3 6 , CFI = .940, and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 129 N F I-.9 8 3 . For Model M l i , %^(22)=30.58, p = .1 0 5 , CFI = .920, and N FI = .975. The two final single group models, and M ^ , are combined to form the initial multiple-group model, MO. Equality of Variable M eans Across Groups In the 1991 Cross-Sectional model, after equality constraints across groups are imposed on measured variables. Model M l has 6 more degrees of freedom than Model MO. For the 1991 Cross-Sectional model, observed variables exhibit mean invariance across groups (M l-M O x^(6)=9.99, p > A). Equality of Factor Loadings Across Groups Model M2, developed from and nested within M odel M l, is used to test whether the factor loadings are equal across the two groups. There are 9 factor loadings in the SEM model, but three of them, one per factor, are fixed at 1 for model identification purposes. M odel M 2, with across-group equality constraints on factor loadings, therefore has 6 additional degrees of freedom over M odel M l. In this case, based on the difference test between M2 and M l, %-(6) = 10.095, p > A, we accept all the constraints on the factor loadings. Equality of Factor Correlations and Regression Weights In the multiple group models previously discussed we have shown partial invariance (equality) of the measurement model across the two Approach Coping groups. M odel M3 imposed a total of 3 additional equality constraints on Model M 2. These constrain the regression coefficients in the model equal across Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 130 Approach Coping groups. The difference test between M odels M3 and M 2, X'(3)=2.96, p>.3, indicates that all of the equality constraints are reasonable. That is, there were no structural differences between the Low and High Approach Coping groups in the 1991 Cross-Sectional model. The final model M3 did not fit the data adequately, x^(57)=80.59, p — .022. However, the N FI and CFI indices, as shown in Thble 14 on page 132, show that the model represents a large portion of the variation in these data (CFI = .903 and N FI = .969). Summary of 1991 Cross-Sectional Longitudinal SEM M odel In summary, analysis of the 1990 Cross-Sectional multiple group structural equation model showed very few differences between groups, even in the measurement model. O f the set of correlated errors added to the two separate single group models before combining them to form the multiple group model, one correlated error was added to both the Low and High Approach Coping groups, three to the High Approach Coping group only, and one to the Low Approach Coping group only, out of a total of 36 possible correlated errors. No significant difference between groups exists in either the measurement or structural portions of the model, except that, as in other models, the mean level of Avoidance Coping was higher in the High Avoidance Coping group. The 1991 Cross-Sectional structural model is summarized in Table 14 on page 132 and in Figure 11 on page 164. This model, which did not control for preexisting differences in the Distress outcome variable, showed support for Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 131 direct positive regression coefficients between Life Events and Avoidance Coping, Life Events and Distress, and Avoidance Coping and Distress. The indirect or mediated effect of Life Events on Distress was significant in the 1991 Cross-Sectional model. As there is only one indirect path in the model, the indirect effect necessarily occurs through mediation by Avoidance Coping. That is, some of the increase in Distress that occurs because of increases in Life Events occurs because increased experience of stressful life events causes increased used of Avoidance Coping, which in turn causes increased psychological distress. The 1991 Cross-Sectional model does not provide good support for the moderator or buffer effect, since no structural differences exist between the Approach Coping groups. S um m ary The five final structural equation models are summarized in Ihble 14 on page 132. Goodness-of-fit indices as well as significant between-group differences and significant indirect effects are shown in the table for each of the five models: Distress Outcome, Fully Lagged, Change Scores, 1990 Cross-Sectional, and 1991 Cross-Sectional. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 132 T able 14. Summary o f Final Two Group Structural Equation M odels XL R N FI CFI Indirect Group Differences Effects Distress Outcome Model 219.30 174 .011 .889 .974 LEg-^D; ACg-^D; (H < L ) SocialCopo (H > L) SocStresS] (H > L ) Fully Lagged Model 309.97 244 .003 .877 .970 LEg-^D; AC q -^D, (H < L )' LEq-^C 1 AntiSocialg (H > L) DoDrugSo (H > L) ACi<»D, (H <L ) Change Scores M odel 146.00 110 .012 .800 .970 None SocStress, (H > L ) Factor Loading o f Stressorsg on ExtStressg (H < L) 1990 Cross-Sectional Model 66.69 54 .1 1 5 .9 2 7 .985 LE q -^Dq SocialCopq (H > L ) 1991 Cross-Sectional M odel 80.59 57 .022 .903 .969 LE,-»D ; No Group Differences Note: In all models except the Change Scores M odel, the mean of Avoidance Coping was greater in the High Approach Coping group. ‘ T his param eter was non-significant in both groups. In the Group Difference column indicates a regression coefficient, "«•” indicates a correlation, other entries are variable means. H > L (or H < L) indicates that the param eter value for the high approach coping group is greater than (or less than) for the low approach coping gro u p , respectively. S ubscripts: 0 = 1990-91, 1 = 1991-92. A bbreviations: L E = L ife Events, A C = A voidance Coping, D =D istress. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 133 D is c u s s io n Overview This study presented findings from a longitudinal, prospective study of a moderately large school-based sample of adolescents in grades 9 through 12 using data collected at two points in time, baseline and 12 month follow-up. M ultiple cross-sectional confirmatory factor analytic models were used to confirm the hypothesized structure of stressor, coping and psychological distress constructs. Longitudinal multiple regression models were used to provide evidence for mediation and moderation of stressor-distress relationships by avoidance and approach coping. Structural equation models were also used to confinn the hypothesized measurement model, consisting of life events, psychological symptoms, and two types of coping: approach and avoidance. Structural equation modeling was also used to correct for the attenuation of direct, mediator, and moderator relationships caused by error in the measurement of the longitudinal model constructs. Stress and Coping Constructs (Measurement Model) Factor analyses confinned the utility of partitioning stressful life events in to multiple domains representing major sources of tension in adolescence. Results suggested that, consistent with developmental theory (e.g. Erikson, 1968), high school students are particularly at risk for conflict and problems from the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 134 interpersonal and employment domains, as their participation in substantive romantic and work relationships is in its formative stages. This study confirmed the separation of coping behaviors into two major domains, termed approach and avoidance coping following Billings, Moos, and others (Billings & Moos, 1982; Holahan & Moos, 1991; M oos et al., 1990). Because of the diversity of coping concepts, literature, and measurement, coping has been difficult to study. Comparisons between the work from different research groups has been hampered by differences in applications of coping theory and measurement. Confinnation of the approach/avoidance coping paradigm in a longitudinal study of adolescents may serve to promote the use of more standard measures of coping. O f the two coping domains identified and validated in this study, approach coping seems to represent more mature and adaptive coping efforts, those that have been labelled “salutary” (Joigensen & Dusek, 1990) because they are hypothesized to be more beneficial to health. These coping strategies attempt to directly and indirectly either alter the stressful situation or change the internal response to the situation. They include utilizing social resources for emotional support or for discussion of problems, engaging in cognitive activities that reduce tension, or engaging in physical activity that can improve self-esteem or productively channel internal conflict. The second type of coping, avoidance coping, focuses on less mature responses to stress or problems including Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 135 antisocial activities such as getting angry, causing trouble, or doing something bad, and avoidance activities including the use of substances (cigarettes, pills, or alcohol), and ignoring the problem by hanging out with friends or going to a party. Stress/Coping/Distress Relationships (Structural Model) Multiple Regression Regression results provide strong support for an adverse effect of proximal life events on psychological distress, consistent with previous work (Coddington, 1972; Johnson & M cCutcheon, 1980; M echanic, 1983; Swearingen & Cohen, 1985). In contrast, support for a direct effect of either approach o r avoidance coping on distress was weak. Only in simple regression models that did not control for any covariates was there a direct effect of Avoidance Coping at Time 0 on Distress at Time 1. The effect lag between coping and distress may be such that baseline avoidance coping has an effect on distress in the immediate post baseline period, which in turn has an effect on distress at follow-up, but these analyses could not confirm that hypothesis. Regression analyses provided some support for mediation by avoidance coping of the stressor-distress relationship. Part of the negative effect of stressful life events seems to occur because of the use of maladaptive coping strategies including aggressive antisocial behavior, drug use and partying. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 136 M ultiple regression with interaction terms confirmed a stress-buffer effect of approach coping, consistent with previous work by others (Cohen & Wills, 1985; Cronkite & Moos, 1984; Wills, 1986; Windle, 1992). A negative regression coefficient for the Approach Coping X Avoidance Coping term means that adolescents who use both approach and avoidance coping techniques have fewer psychological symptoms than those that use only avoidance coping techniques. Structural Equation Models Structural equation models confirmed regression results and clarified relationships after correcting for measurement error. Specifically, cross-sectional models confirmed that relationships between same-time avoidance coping and distress were strong. However, cross-sectional models do not allow easily defensible causal inferences to be made and so longitudinal models were pursued. Longitudinal SEM models also confirmed the conceptual structure of the stressor, distress, approach coping, and avoidance coping constructs previously examined using cross-sectional confirmatory factor analysis. M ore importantly, these SEM models produced results consistent with those from multiple regression while providing additional detail about relationships of interest. As discussed above, the most theoretically consonant model that is easily interpretable is the Distress Outcome model. The cross-sectional models suffer from difficulties of causal attribution although they do provide some insight into Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 137 alternative measurement strategies to capture effects at intervals of less than one year. Of the SEM models tested, the Change Score model is the most difficult to interpret. The Avoidance Coping change score can be quite misleading as it is really a measure of Avoidance Coping stability rather than actual Avoidance Coping. Thus, it equates all adolescents with relatively stable coping repertoires regardless of their level of use of these maladaptive coping strategies and also does not account for momentary or problem-specific transient changes in individual coping strategies. The Fully Lagged model makes maximum use of the constructs measured in the sense that all of the constructs are modeled at both waves, but the added degrees of freedom and the large numbers of parameters to estimate make it empirically a problematic model to work with. In addition, the large amount of the total covariation explained by prediction of follow-up avoidance coping and distress by the baseline levels of these constructs may make the estimation of other parameters of interest more unstable. Common results in the longitudinal SEM models and the regression models include support for a protective effect of approach coping, for a vulnerability or adverse effect of stressful life events, and for significant covariation between approach and avoidance coping. One reason why direct effects from avoidance coping to distress were not apparent may be that adolescents with high levels of avoidance coping tended to have higher levels of approach coping, which masked the negative effect of avoidance coping. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 138 By extending our earlier stress-buffering research (Broder et al., under review) from a cross-sectional to a longitudinal framework, this work supports the finding that stressful life events are associated with increased psychological symptomatology. M ost importantly, we found that the adaptive coping strategies designed to directly and indirectly reduce tension and improve stressful situations can reduce the adverse effects of stressful life events. These relationships hold when controlling for the prior psychological symptoms of anxiety, depression, and somatization. This study also provides support for the proposition that stressful life events predict psychological distress under conditions of both low and high coping resources. Im plications fo r Prevention an d F u tu re R esearch D irections The development of diagnosable psychologic disorders is not a necessary sequelae to the reporting of the psychological symptoms used in this study and the etiology of these disorders is multifactorial and complex. Self-reported psychological symptoms may not be sufficient preconditions for the development of these disorders, but are more likely necessary preconditions. Furthermore, prevention of clinical depression is not the only worthy goal of prevention programs: those programs designed to decrease negative affect in adolescents could also improve their quality of life. Many prior studies have failed to control for baseline levels of outcome variables, so the finding of significant effects of life events and coping strategies on psychological distress at one year while Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 139 controlling for baseline distress represents additional evidence for the importance of life events and coping in the determination of psychological distress. Specifically, the vulnerability factor of stressful life events predicts increased risk for psychological distress. Even stressful life events as distant in tim e as 1-2 years prior to outcome measurement are predictive of depression, anxiety, and somatic symptoms in this sample of adolescents. The predictive value of stressful life events could be used to guide development of population-based prim ary prevention programs to taiget prevention of some of these life events. These programs may be difficult to design as some of the stressful life events seem random and uncontrollable. From a public health perspective, however, many of the stressful life events are preventable. For example, many of the injuries experienced by adolescents are preventable, encompassing homicide, suicide, and motor vehicle accidents. Even events such as “parents got divorced” or “increase in arguments with parents” could be addressed with multi-component community-wide education and behavior modification programs. In the context of the current secular trend emphasizing “family values,” public support for a program addressing family stability and hannony could gam er widespread community support and gain a local champion/sponsor—factors that have been shown to help make previous prevention programs successful (Pentz, M acKinnon, Dwyer, Wang, Hansen, Flay et al., 1989b). Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 140 A secondary prevention strategy would be to use early identification of adolescents at increased risk for psychological symptoms because of the presence of stressful life events as a means of focusing interventions on a subgroup to whom the subject m atter of the prevention program is particularly salient. For example, since moving to a new home and moving to a new school are stressful life events, school districts could identify new students as the taiget population for a prevention program. These students could be provided special resources and education to improve their ability to cope with a stressful new environment. However, there are difficulties inherent in the taigeted population prevention approach. First is the problem of adequately identifying the population of interest. This is essentially a screening problem, where you want both a sensitive and specific method of identifying the high-risk population so that a scarce resource (the prevention program) is allocated to all who need it, but not given by mistake to those who do not. Another problem with “high-risk” programs is that there can be a labelling effect. By labelling adolescents at high-risk for psychological problems there may be an increased tendency for such adolescents to think of themselves as individuals that have problems. Finally, some aspects of the problem are more appropriately addressed by programs that target the entire population. For example, in the case of new students in the school, all students need to be educated about how to welcome the new students. To only taiget the new students would be analogous to teaching African-Americans how to cope Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 141 with racism without attempting to educate the majority population to eliminate racist views. Ultimately, the most productive approach, and one that can reduce the problems of negative labelling is to use both population-based primary prevention and taigeted subgroup prevention. In this “strategic prevention” approach (Pentz & Chou, 1994), the prim ary prevention program sets the stage for the high-risk program. In addition, because psychological distress has been found to be predictive of adolescent substance use (Wills, Vaccaro, & M cNamara, 1992), almost any antecedent of psychological distress could conceivably be a useful taiget of adolescent substance abuse prevention programs. For example, programs teaching conflict-resolution skills could reduce the negative effect of conflicts with parents or with peers, reducing psychological distress and in turn reducing adolescent substance abuse. The effectiveness of population-based approaches has been well- supported (Pentz et al., 1989a; Pentz et al., 1989c). Programs could taiget either the prevention of stressful life events themselves, or a reduction in their negative effect on psychological states. In the latter scenario, our results concerning the vulnerability and protective factors of different methods of coping provide guidance for program development. For example, since a consistent finding was that approach coping methods provide some protection against factors that tend to increase adolescent distress, a program could be designed to teach adolescents to increase the use of these types of coping strategies prior to or during periods of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 142 distress. An issue for further research would be whether such training would be as effective for adolescents who anticipate future psychological distress versus those who do not anticipate distress. Adolescents that anticipate distress may be more aware of their affective states and of their coping responses and so may be better able to benefit from this type of intervention. Furthermore, they also may be more psychologically ready to make an effort to improve their coping skills just as someone who has thought about quitting smoking is more ready to make the actual effort to quit than someone who has never had such thoughts. Such a program should extend work from the more highly developed domains of smoking and drug prevention (Hansen et al., 1988; Pentz, 1994; Pentz et al., 1989a) or from programs aimed at improving cardiovascular health (Perry, Stone, Parcel, Ellison, Nader, Webber et al., 1990). Effective programs are those which have focused on counteracting stressful or negative social influences (Hansen et al., 1988; Pentz, 1994). Specific components of these programs which have enhanced effects include: delivery or program components during at-risk developmental periods such as the year of transition to a new school; use of multiple program components including community- and school-based components; use of peer educators and peer role models; public affirmation of intentions to modify behavior; sufficient program exposure (8-15 school sessions in the first year of intervention); and at least one set of booster sessions in the year following the initial program delivery. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 143 The m ediator results from this study, that the distal adverse effects of stressful life events come at least in part from more proximal effects on distress and not from increased use of avoidance coping strategies, have implications for the design of these prevention programs. Specifically, they suggest focusing on periods of risk defined by high levels of stressful life events, either those that have already occurred or those that can be expected to occur. For example, school transitions are an obvious period of stress in an adolescent’s life, but the period surrounding the first high school dance or the prom are likely also periods of increased stress. Program components could include a social normative component focused on reducing the cognitive appraisal of certain life events as stressful. Such a component could teach adolescents that they are not alone in feeling anxious about the first dance, not everyone has a date already, etc. These messages could be delivered both by peers and through an activity-based program (during athletic-related meetings or during meetings of other clubs in high school). Students could practice acting out situations that they anticipate being stressful such as aiguing with their parents or entering a class for the first time. M oderator results suggest that program components that increase adolescent use of approach coping strategies could provide some protection against increased psychological distress. In this context, also, program components could be delivered both in-school by peers and teachers and out-of- school by popular local celebrities, public service announcements, and through Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 144 places of worship. Students could be encouraged to stand up and promise that the next tim e they have a problem they have to deal with they will use one of the approach coping strategy (preferably they will announce a specific method), rather than an avoidance coping strategy. W hen designing public health interventions, a critical initial decision is whether to taiget the entire population or to use some screening method to identify an at-risk subset to taiget. Economic considerations sometimes dictate the latter, since if screening is inexpensive then more resources can be devoted to helping individuals particularly at-risk if a laige portion of the population is excluded from the intervention based on the screening results. This approach is common in the medical domain and derives from the disease model, where it clearly does not make sense to expend resources treating an entire population if only a subset suffer from the disease being treated. For programs of this type one must often balance the sensitivity and specificity of the screening test used during triage with its cost and the relative resources expended on treatment versus screening. In contrast, a more classic public health perspective is derived from experiences with infectious disease in the last one hundred years, for example with smallpox (Behbehani, 1983), where it often has been necessary to either treat or vaccinate an entire population to successfully solve the public health problem. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 145 The moderator effect findings imply that programs that increase the use of approach coping will be more effective in those adolescents at risk because of increased use of avoidance coping. However, results from the social influences prevention program literature (Hansen et al., 1988; Pentz, 1994) suggest that for behaviors that have a normative component (that is, where a perception of societal encouragement or permissiveness increases the likelihood that an individual will engage in the behavior in question) community- or population-wide programs will be successful. For example, couples at risk for divorce could be identified by a history of previous divorce, history of seeking counseling, or other factors. A program could then intervene to try to provide them with the resources and skills necessary to save their marriages. However, at least part of the reason for a high rate of divorces is the social context in which marriages and interpersonal relationships occur, so that community-wide interventions have their place. For example, one reason that couples may consider divorce is that they believe that their peers approve of divorce as a solution to marital problems. A program which educated the entire community to consider divorce as a solution only after trying other options would focus on changing the normative expectations of individuals. Similarly with the avoidance coping strategies of getting mad, doing something bad, or causing trouble, and the approach coping strategies of playing sports, talking about the problem, or using relaxation techniques. Discouraging the maladaptive strategies and encouraging the adaptive Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 146 ones could be addressed in a program for adolescents who have shown themselves to be at-risk by exhibiting problem behavior. M aking a change in the social climate to make talking about problems socially acceptable and getting mad or causing trouble socially unacceptable could further increase the efficacy of the intervention. Perhaps the best approach is to use a multi-component community-based approach to address a few of the more socially and environmentally motivated coping strategies, but to supplement this with diversion and intervention programs for at-risk adolescents. The community-based intervention should contain components that address areas suggested by both the mediator results (reducing stressors, reducing avoidance coping, and reducing immediate post-stress distress) and the moderator results (increasing approach coping and reducing avoidance coping). The focused intervention should emphasize those behaviors specific to the type of at-risk youth identified, perhaps emphasizing areas suggested by the moderator results of this study. One environment for program delivery is the primary care physician’s office. Adolescents visit their physician’s office for a variety of stated reasons, including to obtain physical exams required for participation in school athletics or required for employment, m inor illness and injuries, preventive medicine needs including PAP smears, treatment for sexually transmitted diseases, and to obtain birth control devices. A productive area for future research would be to examine Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 147 whether a brief office-based intervention could change coping behavior. An appropriate model for how such an intervention would work is the National Cancer Institute’s smoking cessation paradigm for physicians (Glynn & Manley, 1991). The physician’s office may be an appropriate place for an intervention because the piiysician/patient relationship is a unique and powerful one and many adolescents see physicians fairly often. W hat is not yet known is whether physicians can successfully help their patients change their use of coping strategies. Two important characteristics of the N C I’s smoking cessation program is that it does not take a lot of physician time at each encounter and that the entire office acts as a team. There should be literature about the problem-area (stress and coping) freely available in waiting rooms and examining rooms and the office staff other than the physician should be aware that assessing coping behaviors and assisting in changing them if necessary is part of their job. All adolescents should fill out a brief questionnaire about their coping style at their first visit and the physician should follow up on the questionnaire with a brief discussion of the importance of using productive coping strategies and a personalized discussion of whether this patient seems to be using productive (approach) coping strategies or nonproductive (avoidance) coping strategies. If the patient is identified as having a repertoire of coping strategies that should be changed, he should be asked whether he wants to change at every visit. For patients resistant to change, some Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 148 discussion of why change is impoilant and what are common concerns about deciding to change is good, but must be balanced with the amount of time it takes—motivating self-help literature should be provided to these patients. The next step is to assist patients who want to change by giving them specific self- help instructional materials and perhaps setting a particular date to stop using a self-identified maladaptive coping strategy. The final step is to arrange a follow up visit. At the follow up visit the physician should assess the patient’s progress, be encouraging, and be supportive in the case of relapses to maladaptive behavior. The research question to be tested is first, whether physician intervention as described above can change patients’ behavior, and second, whether a program to educate physicians to intervene in this way can produce changes in physician behavior and delivery of a program with adequate fidelity to the one taught to the physicians. In summary, these findings support the utility of primary prevention of adolescent distress through the prevention of those stressful life events for which practical interventions can be developed. In addition, discouraging the use of avoidance coping measures could also be a method for primary prevention of adolescent anxiety, depression, and somatic symptoms. M ulti-component interventions directed at all adolescents with separate components for specific groups of high-risk adolescents could be developed. This might include physician- delivered programs directed at youths identified as being at high risk because of Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 149 the stressful life events of an unintended pregnancy or a sexually transmitted disease. These findings support the teaching of approach coping methods as a method of secondary prevention among adolescents already experiencing psychological distress. Limitations Reliability Internal Consistencv All of the measures used in this study are derived from instruments that have been used with a variety of samples. Because of the limitations on questionnaire size imposed by the nature of the laige community-based study from which these data are taken, the num ber of individual items representing the constructs of interest was less than ideal, typically seven to ten. Nonetheless, the internal consistency of the measures was adequate for the measurement model, ranging from .5 to .9. Both numbers of items used and Cronbach’s alphas for the constructs are shown in Tkble 8 on page 90. Low internal consistency will tend to attenuate results and reduce the power of the study, even with coirection for measurement error as is provided in structural equation modeling. Test-Retest Reliability During a phase of rapid psychosocial development such as adolescence, constructs such as coping are expected to evolve within individuals over time. For this reason measures of test-retest reliability ought to be taken after a short Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 150 enough lag that minimal developmental change will occur. The one year measurement lag in these data is clearly too long for adequate measurement of test-retest reliability in this age group. Across wave correlations for the sub indices are in the .3 range for the stressor indicators, the .6 range for the psychological distress indicators, the .3 to .5 range for the approach coping indicators, and the .4 to .5 range for the avoidance coping indicators. Validity Measurement Construct Validity Issues of measurement construct validity are concerned with whether the constructs are clearly identified and are distinct from one another and whether the measures used accurately reflect the constructs. Many groups of researchers have used the measure of stressors used in this study, stressful life events. As discussed in the background section, other alternatives have been proposed, but have been found wanting in some dimension, particularly with respect to confounding with psychological outcome variables. This is a particularly acute criticism of the Daily Hassles and Uplifts Scale (Dohrenwend et al., 1984; Lazarus et al., 1985) and the Perceived Stress Scale (Cohen et al., 1983). The psychological symptoms measure is a subset of one, the SCL-90-R, that has been shown to accurately reflect psychological symptomatology and to be able to detect depression (Derogatis & Coons, 1993). The SCL-90-R has become so widely accepted that is has been used to validate the construction of newer scales Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 151 (Derogatis, 1990). The coping measures used have also been widely used (Cohen, 1987) and a variety of different coping measures have been shown to have moderate to high correlation with each other (Spirito et al., 1988), showing that there is at least within-construct consistency. The most important aspect of construct validity in studies that make causal attributions is that the relationship ascribed to causality is not due to underlying confounding of constructs. One approach to clarifying the extent to which this is a problem is to use alternative measures of the constructs of interest or to use different outcome constructs altogether. For example. Wills’ (1986) result relating stress and coping to adolescent drug use, and the results of this study relating stress and coping to adolescent psychological distress, together reduce the probability that each suffers from a defect of confounding. Internal Validity of the Study Some threats to internal validity exist in this study. When analysis groups are defined on the basis of a pretest measure containing error, as they are in this study, statistical regression to the mean can become an issue. In the absence of any developmental changes in levels of Approach Coping, the Low Approach Coping group can be expected to have higher follow-up Approach Coping scores than they did at baseline and the High Approach Coping group can be expected to have lower scores than at baseline, raising the question of what the High and Low Approach Coping label means—did the measures adequately identify the Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 152 groups as intended? If part of the reason someone scores high on the Approach Coping items at baseline is because they were particularly self-aware on the day the questionnaire was administered they might also be expected to score high on Avoidance Coping and Distress. If these scores regress to the mean at follow-up a spurious negative regression coefficient might be detected. Similarly, in the absence of a true relationship, analysis of the Low Approach Coping group might produce a spurious positive regression coefficient for the relationship between Avoidance Coping at Distress. Even in the absence of a direct threat to validity, measurement error, one symptom of which is regression to the mean, tends to reduce the power of a study. The major danger of reduced power is to fail to detect a true relationship. From the power discussion above it is clear that this study did not have a laige amount of excess power. If the measurement eiTors are greater than expected, the sample size required for adequate power can increase dramatically—reliability of main effects variables changing from .8 to .7 causes a 50% increase in sample size requirements (Aiken & West, 1991). Given that some interaction effects were detected in this study it is likely that the derived relationships are attenuated and that the true relationships were larger than those used in the power calculations. Aiken and West (1991 p. 167) also point out that measurement error reduces power for detecting interaction effects much more dramatically than it reduces power for detecting main effects. In this way empirical research with Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 153 measures containing error (as do all measures) is inherently biased towards detecting main effects in preference to interaction effects. Otiier possible threats to internal validity relate to reporting bias, adequacy of measures, and the third variable problem. For example, suppose that the coping and distress measures require a certain level of self-awareness to respond accurately to them. Then some of the observed covariation between coping and distress to which we impute causality may merely be a sign of an unmeasured third variable, self-awareness, that causes changes in both predictor and outcome variables. This criticism would fit nicely with the observed significant Avoidance Coping to Distress effect in the High Avoidance Coping group (presumably they are more self-aware) and the lack of effect in the Low Approach Coping group. One way to address this issue is to examine the High Approach Coping group separately and see if differential effects exist within the group. External Validity of the Findings Small sample sizes when these data were stratified by gender or age precluded the use of SEM to examine the structural model in these subgroups. As expected, mean levels of the factors did not differ significantly by age. However, also as expected, females reported more stressful life events and more psychological distress than males. It is well known that females report more psychological symptoms and that they are diagnosed with certain psychological illnesses (such as depression) more frequently than men. This study does not Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 154 clarify the specific import of the observed mean differences. Rather, it focused on measuring stress relationships in whole, representative adolescent populations with grade, gender, and socioeconomic status fairly balanced. However, since the majority of the study analyses grouped both genders and all three grades together and found significant effects, structural differences between groups must not be too profound. M ore precisely, it seems likely that at least the direction of relationships, positive or negative, holds true across gender and grade lines, if not equality of magnitude of the relationships. The instruments used in this work were modified specifically for adolescent populations, but many stressor-coping- distress studies have been performed on adult populations, making it more likely that these results are generalizable. These analyses grouped together students that received a drug, alcohol, and tobacco prevention intervention (the program group) and those that did not (the control group). Because the intervention tested in the overall study did not focus specifically on changing coping behavior program and control groups were combined for analysis. This provided additional statistical power by increasing sample sizes as well as contributing to increased generalizability of results. In conclusion, future directions for research include an examination of alternative measurement periods to address the issue of effect lag; comparison of the coping measures used in this study with alternatives so that ultimately the coping measurement field can be more unified; replication of this study with a Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 155 larger sample and more reliable (lengthier) measures to clarify the significance of some of the marginal effects detected; and most importantly, use of a true experimental design with an intervention formulated to manipulate coping behavior so that strong causal inferences about observed results can be made. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■ D O Q. C g Q. $ I —H 3" m Table 15. Descriptive Statistics and Correlations of Measured Variables (SEM Model Input) by Approach Coping Group 3 Adaptive Coping High (N=206) ^ Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 o ' Mean -.15 -.05 -.00 .17 . 04 .17 .05 .02 .06 .43 .42 .50 -.26 -.12 .09 .10 .06 .09 .09 .07 .05 .26 .25 .25 o SD 1.85 1.57 2.74 .75 .97 . 87 .77 .79 . 80 .86 .70 .79 2.14 1.58 2.84 .76 .87 . 86 .81 .88 .83 .90 .76 .90 Z: Skew 1.49 1.92 .58 .85 2.37 . 37 .49 1.66 1.00 .12 .07 .29 2.15 1.39 .35 1.00 2.19 .45 .41 1.66 1.31 .31 .09 .49 m Kurt 1.822.81 -.26 .5 75.16 -.72 .122.83 .61 -.88 -.47 -.33 3.81 .69 -.80 1.16 5.09 -.43 -.10 2.84 1.52 -.77 -.51 -.35 g 1. Extstro -.03 2.15 1.82 2.95 .24 .49 -.10 .00 .12 -.17 .01 -.02 -.04 -.12 .04 .23 -.08 -.01 .06 .06 -.00 -.10 -.05 .02 .14 .08 .08 ;o 2. Jobstro .05 1.65 1.75 2.04 .02 .22 .02 . 09 .10 .02 .03 -.03 -.10 -.01 -.07 .11 .31 .04 .08 .22 .12 -.01 .06 -.00 .03 -.04 .01 (§■ 3. Soostro -.09 3.03 .77 -.06 .26 .21 .25 .14 .12 .21 .34 .29 -.10 .09 .14 .05 .08 .33 .24 .19 .04 .11 .14 .21 .04 .10 .08 3 : 4. A n tiso o "-.16 . 60 .76 .48 -.07 . 06 .22 . 39 .11 .59 .45 .41 .13 .18 .19 .04 .07 .29 .49 .26 .02 . 35 .22 .19 -.08 .03 .12 0 5. Dodrugo -.04 . 69 2.07 3.64 . 05 .06 .15 .54 . 30 .24 .26 . 30 .03 .02 .10 .01 .00 .14 .11 .52 .06 .14 .10 .18 -.12 .00 .02 1 6. Socialo” -.18 . 86 .46 -.59 . 04 . 01 .08 .21 .32 .02 -.03 .12 .07 -.01 .10 -.08 -.06 -.04 .04 .16 .46 .00 -.09 .07 -.00 .15 .04 g 7. Anxietg -.07 .74 .35 .14 -.00 .03 .37 .44 .18 .28 .59 .54 .01 .29 .20 -.03 .13 .31 .33 .12 .12 .53 .32 .37 -.11 .09 .05 8. Depreso -.04 .70 1.32 1.34 -.06 . 05 . 32 .51 .33 .03 . 54 . 52 -.06 .21 .28 .10 .11 .41 .24 .21 .05 .41 .55 . 39 -.09 -.02 .06 c 9. Somati(,+ -.07 .72 .96 .48 .02 -.10 .37 .36 .22 .10 .58 .58 -.08 .30 .21 -.00 .09 .37 .28 .17 .07 .42 .31 .57 -.10 .16 .10 3- 10. Exercio" -.48 .64 . 87 . 61 -.03 .03 .02 .14 .14 .17 -.02 -.05 -.08 -.15 -.05 .11 -.11 -.04 .13 -.01 -.05 -.04 -.04 -.07 .41 -.07 -.04 ^ 11. Getinfo" -.45 . 51 .43 -.06 -.08 -.10 -.13 -.00 . 01 -.00 -.18 -.20 -.10 - . i l .16 -.13 .11 .14 .09 -.09 .17 .25 .12 .13 -.11 .34 .12 S 12. Relaxo" -.55 . 52 .48 -.13 -.03 -.07 -.07 .05 .17 .13 .11 .05 .12 -.17 -.02 .00 -.10 .10 .13 .07 .08 .19 .15 .15 -.08 .19 .29 3 13. Extstr, .11 2.42 1.70 1.92 .22 . 04 .17 . 08 .10 .04 .04 .12 .07 . 07 -.13 -.10 .26 . 33 .03 .03 -.12 .09 .22 .19 .02 -.02 -.04 g- 14. Jobstr, .07 1.81 1.15 -.27 .06 .21 .08 .08 -.03 -.02 .10 .08 .06 -.05 -.04 .11 .28 .34 .06 .09 -.01 .15 .15 .21 -.07 .02 -.04 a 15. Socstrj -.17 2.98 .56 -.49 .06 .07 . 36 .22 .10 .16 .22 .22 .21 .03 -.10 .02 .24 .21 .36 .22 .18 .38 .45 .45 -.05 .06 .09 ° 16. Antisoi" -.08 . 63 .69 1.32 .03 -.01 .08 .47 . 36 .17 .22 .28 .24 . 00 . 01 .18 .13 .14 .09 . 31 .18 .44 .26 .23 .10 .14 .30 ■a 17. Dodrugi -.04 .72 1.78 2.38 .05 .06 .07 .25 .41 .24 .11 .24 .09 -.01 -.01 .10 .16 .02 .07 .49 .09 .10 .17 .14 .00 .04 .09 o 18. Social,' -.10 . 84 .55 -.00 -.03 -.01 .04 .08 .10 .41 .20 .07 .13 .02 .09 . 08 .00 .08 .14 .20 .28 .18 .01 .14 .15 .27 .26 C T 19. A nxiet,' -.06 .75 . 31 -.04 -.01 -.03 .23 .41 .13 .15 .53 .40 .37 -.01 -.13 .17 -.03 .04 .27 .39 .10 .03 .58 .53 -.07 .05 .06 3 20. Depres, -.04 .75 1.72 3.89 .01 -.02 .26 . 36 .22 .12 . 39 .54 .38 .05 -.19 .12 .07 .02 .30 .46 .29 .01 .62 .48 .02 -.10 .12 21. Somali, -.00 . 81 1.28 1.82 .05 -.12 .28 . 30 .14 .09 .44 .42 .55 -.04 -.07 .05 -.01 .03 .27 . 33 .12 .01 .58 .64 -.09 .12 .06 22. E x erci," -.24 . 84 . 88 -.01 -.09 -.04 -.02 -.00 -.01 .04 -.01 -.04 -.00 .41 -.02 -.06 -.09 -.10 -.05 .06 .01 .11 .03 -.04 -.00 . 32 . 33 o 23. G etin f," -.21 .65 .46 .01 .01 .04 . 04 -.14 -.05 . 04 -.02 - .i l .01 -.13 .41 -.04 -.08 -.02 -.12 .03 -.08 .23 -.12 -.25 -.06 .10 . 38 ^ 24. R elax ," -.22 .72 . 61 -.04 -.06 -.03 -.01 -.07 -.02 . 08 .04 .03 .09 -.14 .20 .24 -.09 -.03 .00 .21 .15 .26 .11 .10 .11 .22 .40 C D Adaptive Coping Low (N=195) Q. C /) C /) Note: r ’s h .l4 or i-.1 4 are significant at p < .0 5 , for both groups. ^ = p < .1 0 , * = p < .0 5 , " = p< .0001 for group mean differences. L n O N 157 Table 16. Summary o f Items Used in Subscale Construction COPING Below is a list of things that people do when they have a problem. For each one, check HOW OFTEN you respond or act that way. (1 = Almost Never, 2 = Sometimes, 3 = Often, 4 = Almost Always) Approach Coping Exercise I go out and play sports I work it o ff by physical exercise Getinfo I talk with my mother or father 1 talk with one of my friends I get information that is needed to deal with the problem Relax I try deep breathing 1 try to calm myself PSYCHOLOGICAL DISTRESS Below is a list o f problems that people sometimes have or feel. For each one, please check HOW OFTEN each problem happens to you. (1—Almost Never, 2=Som etim es, 3 = Often, 4 = Almost Always) Depression I have been seriously thinking about a way to hurt myself I feel like I have nothing to look forward to 1 am no good for anything at all I feel lonely I feel sad Anxiety I worry about things in my life I feel like 1 don’t want to do any tiling I find it hard to keep my mind on something I’m working on I am tired during the day Somatic Svmptoms I feel dizzy or light headed I get headaches I have pains in my heart or chest Avoidance Coping AntiSoc I get mad at people I do something bad or cause trouble I sleep more I get away from people DoDrugs 1 smoke a cigarette to relax I take pills (like aspirin tranquilizers) to feel better I drink alcohol Social I hang out with friends I go to or have a party STRESSORS Below is a list o f events that sometimes happen to people your age. For each of the events check if tlie event happened to you in the last 12 MONTHS. (l= Y e s , 0 = N o) Scale Item Ext Stress Moved to new home Ext Stress Changed to new school Ext Stress Parents divorced or separated or Job Stress Changed to new job Job Stress Quit job Soc Stress Lost a close friend Soc Stress Increase in # of arguments with parents Soc Stress Major illness or injury to yourself Soc Stress Broke up with boyfriend/ girlfriend Soc Stress Gained or lost a lot of weight Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. C D ■D O Q. C g Q. ■ D C D C/) o' 3 8 ■ D Distress Outcome Multiple Group Structural Model (Unstandardized) 3. 3 " C D C D ■ D O Q . C a O 3 ■ D O C D Q. Life Events A v o id a n c e Coping 90 Distress 91 Distress 90 Life Events Life Events -.305 (ns) Avoidance Coping 90 .077 3.704 .088 .094 Distress 91 .165 .191 Distress 90 .746 .134 3.179 .092 Life Events 7.559 Note: The figure on the left is the full model as tested. Only the structural model is shown. Values are unstandardized parameter estimates. The figure on the right is the final multiple group model with only paths and model parameters significant at p<.10 shown. Paths significant at p<.10 but >.05 are shown with a dashed or grayed line, those significant at p<.05 are shown in black. Values within circles are variances. Double-headed arrows represent correlations (values are covariances), single-headed arrows represent regression weights. If a single parameter is shown, then there were no significant group differences. If two parameters are shown, the first parameter is for tlie High Approach Coping (High Positive Coping) group, with the parameter for the Low Approach Coping group shown in parentheses. Ns indicates that the parameter was not significantly different from 0. Small circles indicate disturbances. T3 C D The only group differences not shown arc differences on the means of two measured variable and one construct: the means of Social Coping at Time 0 and Social Life Events at Time 1 and Avoidance Coping at Time 0 were higher in the High Approach Coping Group. Statistically significant indirect (mediated) effects include the effect of Life Events 90 on Distress 91. (/) (/) L n 00 C D ■D O Q. C g Q. ■ D C D C/) C/) 8 ■ D ( O ' 3. 3" C D C D ■ D O Q . C a O 3 "O O C D Q . Life Events 90 Avoidance Copine 90 Distress 91 Distress 90 Life Events 91 Distress Outcome Multiple Group Structural Model (Standardized) -.185 (ns) Life Events A v o id a n c e Coping 90 .448 .894 .730 Distress 91 .589 .808 Distress 90 .670 .737 .601 .462 Life Events Note: The figure on the left is the full model as tested. Only the structural model is shown. Values arc standardized parameter estimates. The figure on the right is the final multiple group model with only paths and model parameters significant at p<. 10 shown. Paths significant at p<. 10 but >.05 are shown with a dashed or grayed line, those significant at p<.05 are shown in black. Values within circles are variances. Double-headed arrows represent correlations, single-headed arrows represent regression weights. If a single parameter is shown, then there were no significant group differences. If two parameters arc shown, the first parameter is for the High Approach Coping group, with the parameter for the Low Approach Coping group shown in parentheses. Ns indicates that the parameter was not significantly different from 0. Small circles indicate disturbances. The only group differences not shown are differences on the means of two measured variables and one construct: the means of Social Coping at Time 0 and Social Life Events at Time 1 and Time 0 Avoidance Coping were higher in the High Approach Coping Group. The indirect (mediated) effect of Life Events 90 on Distress 91 is statistically significant. ■ D C D (/) (/) V O C D ■ D O Q. C g Q. ■ D C D C / ) C / ) 8 Fully Lagged Multiple Group Structura! Model (Unstandardized) 3 3" ( D ( D T3 O Q. C a o 3 T3 O ( D Q. T3 ( D .526 Life Events .110 Avoidance Coping 90 Avoidance Coping 91 .162 Life Events Avoidance Coping 91 .■\voidancc Coping 90 3.229 .041 .091 ns (.071) .184 .186 .126 .035 Distress 90 Distress 91 Distress 91 Distress 90 .705 3.159 .082 Life Events Life Events 1.542 Note: The figure on the left is the full model as tested. Only the slrucltiral model is shown. Values are unslandardized parameter estimates. The figure on the right is the final multiple group model with only paths and model parameters significant at p<.10 shown. Paths significant at p<.10 but >.05 are shown with a dashed or grayed line, those significant at fx.05 are shown in black. Values within circles are factor variances. Double-headed arrows represent correlations (values are covariances), single-headed arrows represent regression weights. If a single parameter is shown, then there were no significant group differences. If two parameters are shown, the first parameter is for the High Approach Coping (High Positive Coping) group, with the parameter for the Low .Approach Coping group shown in parentheses. Ns indicates that the parameter was not significantly different from 0. Small circles are disturbances. The only group differences not shown are differences on the means of two measured variables and one construct: the means of Antisocial Coping, Do Drugs Coping, and Avoidance Coping all at Time 0 were higher in the High Approach Coping Group. The regression weight between Avoidance Coping 90 and Distress 91 was significantly higher in the Low Approach Coping Group, but in neither was significantly different from 0. Statistically significant indirect (mediated) effects include the effect of Life Events 9 0 on Distress 91 and the effect of Life Events 90 on Avoidance Coping 91. (/) (/) g C D ■ D O Q. C g Q. ■ D C D C/) C/) 8 ■ D C D 3. 3 " C D C D ■ D O Q. C a O 3 "O O C D Q. ■ D C D C /) C /) Fully Lagged Multiple Group Structural Model (Standardized) .670 Life Events Avoidance Coping 90 Avoidance Coping 91 .877 .481 Life Events Avoidance Coping 91 Avoidance Coping 90 .641 ns (.4) .616 .782 .347 Distress 90 Di.stre.ss 91 Distress 91 Distrc.ss 90 .705 .601 .442 Life Events Life Events Note: The figure on the left is the full model as tested. Only the structural model is shown. Values are standardized parameter estimates. The figure on the right is the final multiple group model with only paths and model parameters significant at p<.-0 shown. Paths significant at p<. 10 but >.05 are shown with a dashed or grayed line, those significant at p<.05 are shown in black. Values within circles are factor variances. Double-headed arrows represent correlations (values are covariances), single-headed arrows represent regression weights. If a single parameter is shown, then there were no significant group differences. If two parameters are shown, the first parameter is for the High Approach Coping (High Positive Coping) group, with the parameter for the Low Approach Coping group shown in parentheses. Ns indicates that the parameter was not significantly different from 0. Small circles are disturbances. The only group differences not shown are differences on the means of two measured variables and one construct: the means of Antisocial Coping, Do Drugs Coping, and .Avoidance Coping all at Time 0 were higher in the High Adaptive Coping Group. The regression weight between ■A voidance Coping 90 and Distress 91 was significantly higher in the High Adaptive Coping Group, but in neither was significantly different from 0. Statistically significant indirect (mediated) effects include the effect of Life Events 90 on Distress 91 and the effect of Life Events 90 on Avoidance Coping 91. C D ■D O Q. C g Q. ■ D C D C / ) C / ) Change Scores Multiple Group Structural Model C D 8 3 3" ( D ( D T3 O Q. C a o 3 T3 O Life Evcnis I Change in Avoidance Y Coping Change in Distress Life Evcnis -.033 Life Events Avoidance Coping -W Distress .671 9.814 .082 .051 2.124 .182 Life Events 6.151 Note; The figure on the left is the full model as tested. Only the structural model is shown. Avoidance Coping and Distress are shown without a year specified because they are change scores: value in 1991-2 minus value in 1990-1. Values are unslandardized parameter estimates. The figure on the right is the final multiple group model with only paths and model parameters significant at p<.10 shown. Paths significant at p<.10 but >.05 are shown with a dashed or grayed line, those significant at p<.05 are shown in black. Values within circles are factor variances. Double-headed arrows represent correlations (values are covariances), single-headed arrows represent regression weights. If a single parameter is shown, then there were no significant group differences. If two parameters are shown, the first parameter is for the High Approach Coping (High Positive Coping) group, with the parameter for the Low Approach Coping group shown in parentheses. Ns indicates that the parameter was not significantly different from 0. Small circles are disturbances. C D Q. The only group differences not shown are differences on the means of one measured variable and on one factor loading; the mean of Social Life Events at Time 1 was higher in the High Approach Coping Group and the factor loading of Life Events 90 on External Stressors 90 was higher in the Low Approach Coping Group. There were no statistically significant indirect (mediated) effects. T3 C D (/) (/) 8 C D " D O Q. C g Q. ■ D C D C/) C/) 1990 Cross-Sectional Multiple Group Structural Model 8 "O 3 C D .066 Life Events Avoidance Coping 00 Life Evcnis Avoidance Coping 90 Distress 90 .045 Distress 90 .747 6.850 .096 .170 3. 3 " C D C D " O O Q. C a O 3 ■ D O C D Q. Note: The figure on the left is the full model as tested. Only the structural model is shown. Values are unstandardized parameter estimates. The figure on the right is the final multiple group model with only paths and model parameters significant at p<.10 shown. Paths significant at p<.10 but >.05 are shown with a dashed or grayed line, those significant at p<.05 arc shown in black. Values within circles are factor variances. Double-headed arrows represent correlations (values arc covariances), single-headed arrows represent regression weights. If a single parameter is shown, then there were no significant group differences. If two parameters are shown, the first parameter is for the High Approach Coping (High Positive Coping) group, with the parameter for the Low Approach Coping group shown in parentheses. Ns indicates that the parameter was not significantly different from 0. Small circles arc disturbances. The only group differences not shown are differences on the means of one measured variables and one construct: the mean of Social Coping and Avoidance Coping at Time 0 were higher in the High Approach Coping Group. Statistically significant indirect (mediated) effects include the effect of Life Events 90 on Distress 90. ■ D C D (/) (/) a C D ■ D O Q. C g Q. ■ D C D C /) C /) 1991 Cross-Sectiorial Multiple Group Structural Model 8 ■ D .0 8 0 L ife E v e n ts A v o id a n c e C o p in g VI L ife E v e n ts A v o id a n c e C o p in g 91 - W D is tr e s s 91 D is tr e s s 9 1 .0 5 5 .5 9 3 8 .8 9 0 .059 .225 3. 3 " C D C D "O O Q. C a O 3 "O o Note: The figure on the left is the full model as tested. Only the structural mode! is shown. Values are unstandardized parameter estimates. The figure on the right is the final multiple group model with only paths and model parameters significant at p<.10 shown. Paths significant at p<.10 but >.05 are shown with a dashed or grayed line, those significant at p<.05 are shown in black. Values within circles are factor variances. Double-headed arrows represent correlations (values are covariances), single-headed arrows represent regression weights. If a single parameter is shown, then there were no significant group differences. If two parameters are shown, the first parameter is for the High Approach Coping (High Positive Coping) group, with the parameter for the Low Approach Coping group shown in parentheses. Ns indicates that the parameter was not significantly different from 0. Small circles arc disturbances. C D Q. The only group difference not shown is a difference on the mean of one construct: the mean of Time 1 Avoidance Coping was higher in the High Approach Coping Group, Statistically significant indirect (mediated) effects include the effect of Life Events 91 on Distress 91. ■ D C D C /) C /) 165 R e f e r e n c e s Aiken, L. S., & West, S. G. (1991). M ultiple regression: Testing and interpreting interactions. Newbury Park, CA: SAGE Publications. Alexander, F. (1950). Psychosomatic medicine, its principles and applications (1st ed.). New York: Norton. Ander, S., Lindstrom, B., & Tibblin, G. (1974). Life changes in random samples of middle-aged men. In E. K. E. Gunderson & R. H. Rahe (Eds.), Life stress and illness (pp. 121-124). Springfield, IL: Charles C. Thomas. Antoni, M . H ., LaPerriere, A ., Schneiderman, N ., & Fletcher, M . A. (1991). Stress and immunity in individuals at risk for AIDS. Stress M edicine. 7(1), 35-44. Antoni, M . H ., Schneiderman, N ., Fletcher, M . A ., Goldstein, D. A ., Ironson, G., & Laperriere, A. (1990). Psychoneuroimmunology and HIV-1. Journal of Consulting and Clinical Psychology. 58(1). 38-49. Antonovsky, A. (1979). Health, stress, and coping. San Francisco: Jossey-Bass. Archer, S. L ., & Waterman, A. S. (1994). Prim ary prevention of adolescent drug abuse. In C. B. Fisher & R. M. Lem er (Eds.), Applied Developmental Psychology (pp. 76-100). New York: M cGraw-Hill, Inc. Baron, R. M ., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology. 51(6), 1173-1182. Bauman, K. E ., & Dent, C. W. (1982). Influence of an objective measure on self-reports of behavior. Journal of Applied Psychology. 67(5), 623-628. Beck, A. T , Ward, C. H ., M endelson, M ., M ock, J., & Erbaugh, J. (1961). An inventory for measuring depression. Archives of General Psychiatry. 4, 561-571. Behbehani, A. M. (1983). The smallpox story: life and death of an old disease. Microbiology Reviews. 47(4), 455-509. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 6 Bentler, P. M . (1987). Drug use and personality in adolescence and young adulthood: Structural models with nonnormal variables. Child Development. 58. 65-79. Bentler, P. M . (1989). EOS structural equations program m anual. Los Angeles, CA; BM DP Statistical Software, Inc. Bentler, P. M. (1990a). Comparative fit indexes in structural models. Psvchological Bulletin. 107(2), 238-246. Bentler, P. M . (1990b). Fit indexes, Lagrange multipliers, constraint changes and incomplete data in structural models. M ultivariate Behavioral Research. 25(2), 163-172. Bentler, P. M ., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin. 88. 588-606. Bentler, P. M ., & Chou, C P. (1987). Practical issues in structural modeling. Sociological M ethods and Research. 16(1). 78-117. Bentler, P. M ., & Chou, C.-P. (1992). Some new covariance structural model improvement statistics. Sociological Methods and Research. 21(2), 259-282. Bernard, C. (1858). Leçons sur la physiologie et la pathologie du svsteme nerveux. Paris, France: J.B. Bailliere. Bernard, C. (1927). An introduction to the study of experimental medicine (Greene, H enry Copley, Trans.). New York: The M acM illan Co. Billings, A. C., & Moos, R. H . (1982). Stressful life events and symptoms: A longitudinal model. Health Psychology. 1(2), 99-117. Billings, A. G., & Moos, R. H. (1981). The role of coping responses and social resources in attenuating the stress of life events. Journal of Behavioral M edicine. 4, 139-157. Blaney, N. T., Goodkin, K ., Morgan, R. Q , Feaster, D., M illon, C., Szapocznik, J., & Eisdorfer, C. (1991). A stress-moderator model of distress in early HIV-1 infection: concurrent analysis of life events, hardiness, and social support. Journal of Psychosomatic Research. 35(2/3), 297-305. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 167 Blaney, P. H. (1985). Stress and depression in adults: A critical review. In T. M. Field, P. M . M cCabe, & N. Schneiderman (Eds.), Stress and Coping Hillsdale, NJ: Lawrence Erlbaum Associates. Borenstein, M ., & Cohen, J. (1988). Statistical power analysis: A computer program. In Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Botvin, G. J. (1986). Substance abuse prevention research: recent developments and future directions. Journal of School H ealth. 56. 369-374. Botvin, G. J., Baker, E ., Dusenbury, L ., Tortu, S., & Botvin, E. M . (1990). P rev en tin g ad o lesc en t d rug ab u se th ro u g h a m u ltim o d al cognitive-behavioral approach: results of a 3-year study. Journal of Consulting and Clinical Psvchologv. 58(4), 437-46. Brantley, P. I , Waggoner, C. D., Jones, G. N ., & Eappaport, N. B. (1987). A daily stress inventory: Development, reliability, and validity. Journal of Behavioral M edicine. 10(1), 61-74. Broder, B. I., Pentz, M . A ., & Chou, C.-P. (under review). M ediating and moderating effects of coping skills on adolescent stress. . Brown, S. A. (1989). Life events of adolescents in relation to personal and parental substance abuse. American Journal of Psvchiatrv. 146. 484-489. Cannon, W. B. (1915). Bodilv changes in pain, hunger, fear, and rage: an account of recent researches into the function of emotional excitement. New York: D. Appleton and Co. Chou. C .-P , & Bentler, P. M . (1990). M odel modification in covariance structure modeling: A comparison among likelihood ration, Lagrange multiplier, and Wald tests. Multivariate Behavioral Research. 25(1), 115-136. Coddington, R. D. (1972). The Significance of Life Events As Etiologic Factors in the Disease of Children-H A Study of a Normal Population. Journal of Psvchosomatic Research. 16. 205-213. Cohen, F. (1987). Measurement of coping. In S. V. Kasl & C. L. Cooper (Eds.), Stress and health: Issues in research methodology, (pp. 283-305). Chichester, England: John Wiley and Sons. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 1 6 8 Cohen, S., Kamarck, T., & M ermelstein, R. (1983). A global measure of perceived stress. Journal of Health and Social Behavior. 24. 385-396. Cohen, S., Tyrrell, D. A. J., & Smith, A. R (1991). Psychological stress and the common cold. New England Journal of M edicine. 325. 606-612. Cohen, S., & Wills, T. A. (1985). Stress, social support, and the buffering hypothesis. Psvchological Bulletin. 98(2), 310-357. Comfrey, A. L. (1978). Common methodological problems in factor analytic studies. Journal of Consulting and Clinical Psvchologv. 46(4), 648-659. Compas, B. E. (1987). Coping with stress during childhood and adolescence. Psvchological Bulletin. 101(31. 393-403. Compas, B. E ., M alcame, V. L ., & Fondacaro, K. M. (1988). Coping with stressful events in older children and young adolescents. Journal of Consulting and Clinical Psvchologv. 56(3), 405-411. Compas, B. E., Wagner, B. M ., Slavin, L. A ., & Vannatta, K. (1986). A prospective study of life events, social support, and psychological symptomatology during the transition from high school to college. American Journal of Community Psvchologv. 14(31. 241-257. Cook, T. D., & Campbell, D. T. (1979). Quasi-Experimentation: Design & Analysis Issues for Field Settings. Boston, MA: Houghton M ifflin Co. Coulehan, J. L ., Schulbeig, H. C., & Block, M . R. (1989). The efficiency of depression questionnaires for case finding in primary medical care. Journal of General Internal M edicine. 4, 542-547. Coyne, J. C., Aldwin, C., & Lazarus, R. S. (1981). Depression and coping in stressful episodes. Journal of Abnormal Psvchologv. 90. 439-447. Cronkite, R. C., & Moos, R. H. (1984). The role of predisposing and moderating factors in the stress-illness relationship. Journal of Health and Social Behavior. 25(December), 372-393. Depression Guideline Panel (1993). Depression in Primary Care: Volume 1. Detection and Diagnosis. Clinical Practice Guideline. Num ber 5. No. AHCPR Publication No. 93-0550). U.S. Department of Health and Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 169 Human Services, Public Health Service, Agency for Health Care Policy and Research. Derogatis, L. R. (1975). The SCL-90-R. Baltimore: Clinical Psychometric Research. Derogatis, L. R. (1982). Self-report measures of stress. In L. Goldbeiger & S. Breznitz (Eds.), Handbook of Stress: Theoretical and Clinical Aspects (pp. 270-294). New York: The Free Press. Derogatis, L. R. (1990). SCL-90-R: A bibliography of research reports. Baltimore: Clinical Psychometric Research. Derogatis, L. R ., & Coons, H. L. (1993). Self-report measures of stress. In L. Goldbeiger & S. Breznitz (Eds.), Handbook of Stress: Theoretical and Clinical Aspects (pp. 200-233). New York: The Free Press. Derogatis, L. R ., Lipman, R. S., Rickels, K ., Uhlenhuth, E. H ., & Covi, L. (1974). The Hopkins Symptom Checklist (HSCL): A Self-report symptom inventory. Behavioral Science. 19, 1-15. Dohrenwend, B. S., & Dohrenwend, B. P. (Ed.). (1974). Stressful life events: their nature and effects. New York: Wiley. Dohrenwend, B. S., Dohrenwend, B. P., Dodson, M ., & Shrout, P. E. (1984). Symptoms, hassles, social supports, and life events: Problem of confounded measures. Journal of A bnonnal Psvchologv. 93(2), 222-230. Dryfoos, J. G. (1990). Adolescents at risk. New York: Oxford University Press. Dwyer, J. H. (1983). Statistical models for the social and behavioral sciences. New York: Oxford University Press, Inc. Dwyer, J. H ., M acKinnon, D. P., Pentz, M. A ., Flay, B. R ., Hansen, W. B., Johnson, C. A ., & Wang, E. Y.-I. (1989). Estimating intervention effects in longitudinal studies: The Midwestern Prevention Project. American Journal of Epidem iology. 130. 781-795. Ebata, A ., & Moos, R. H . (1991). Coping and adjustment in distressed and healthy adolescents. Journal of Applied Developmental Psvchologv. 12. 33-54. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 170 Eitinger, L. (1973). M ortality and morbidity after excessive stress. A follow-up investigation of Norwegian concentration camp survivors. New York: Humanities Press. Eitinger, L ., & Major, E. F. (1993). Stress of the Holocaust. In L. Goldbeiger & S. Breznitz (Eds.), Handbook of Stress: Theoretical and Clinical Aspects (pp. 617-640). New York: The Free Press. Elliott, G. R ., & Eisdorfer, C. (Ed.). (1982). Stress and human health: analysis and implications of research: a study by the Institute of Medicine. National Academv of Sciences. New York: Springer Publishing CO. Epstein, L. H ., & Perkins, K. A. (1988). Smoking, stress, and coronary heart disease. Journal of Consulting and Clinical Psychology. 56(3), 342-349. Erikson, E. H. (1963). Youth: Change and challenge (1st ed.). New York: Basic Books. Erikson, E. H. (1968). Identitv: Youth and crisis. New York: Norton. Evans, R. I., Hansen, W. B., & M ittelmark, M . B. (1977). Increasing the validity of self-reports of smoking behavior in children. Journal of Applied Psvchologv. 62, 521-523. Folkman, S., & Lazarus, R. S. (1980). An analysis of coping in a middle-aged community sample. Journal of Health and Social Behavior. 21 (September). 219-239. Folkman, S., & Lazarus, R. S. (1986). Stress processes and depressive symptomatology. Journal of Abnormal Psvchologv. 95, 107-113. Glaser, R ., & Kiecolt-Glaser, J. (1988). Stress-associated immune suppression and acquired immune deficiency syndrome (AIDS). In T. P. Bridge, A. F. Mirsky, & F. K. Goodwin (Eds.), Psychological, neuropsychiatrie, and substance abuse aspects of AIDS (pp. 203-215). New York: Raven Press. Glynn, T. J., & Manley, M . W. (1991). How to help your patients stop smoking: A National Cancer Institute manual for phvsicians No. 93-3064). U.S. Departm ent of Health and Human Services, Public Health Service, National Institutes of Health. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 171 Gorman, J. M ., & Kertzner, R. (1990). Psychoneuroimmunology and HIV infection. Journal of Neuropsychiatry and Clinical Neurosciences. 2, 241-252. Graham, J. W., Flay, B. R ., Johnson, C. A ., Hansen, W. B., Grossman, L ., & Sobel, J. L. (1984). Reliability of self-report measures of drug use in prevention research; Evaluation of the Project SMART Questionnaire via the test-retest reliability matrix. Journal of Drug Education. j4 (2 ), 175-193. Hansen, W. B., Johnson, C. A ., Flay, B. R ., Graham, J. W., & Sobel, J. (1988). Affective and social influences approaches to the prevention of multiple substance abuse among seventh grade students: results from Project SMART. Preventive M edicine. 17, 135-154. Hawkins, W. E ., Hawkins, M. J., & Seeley, J. (1992). Stress, health-related behavior and quality of life on depressive symptomatology in a sample of adolescents. Psvchological Reports. 71(1), 183-6. Hinkle, L. E ., Jr. (1977). The concept of "stress" in the biological and social sciences. In Z. J. Lipowski, D. R. Lipsitt, & P. C. Whybrow (Eds.), Psvchosomatic medicine: Current trends and clinical implications (pp. 27-49). New York: Oxford University Press. Holahan, C. J., & Moos, R. H. (1991). Life stressors, personal and social resources, and depression: a 4-year structural model. Journal of Abnonnal Psvchologv. 100(1), 31-38. Holmes, T. H ., & Rahe, R. H. (1967). The Social Readjustment Rating Scale. Journal of Psvchosomatic Research. JJ,, 213-218. Horowitz, M . J., Krupnick, J., Kaltreider, N ., Wilner, N ., Leong, A ., & M anner, C. (1981). Initial psychological response to parental death. Archives of General Psvchiatrv. 38. 316-323. Horowitz, M. I , Schaefer, C., & Cooney, P. (1974). Life event scaling for recency of experience. In E. K. E. Gunderson & R. H . Rahe (Eds.), Life stress and illness (pp. 125-133). Springfield, IL: Charles C. Thomas. Irwin, M ., Daniels, M ., Bloom, E. T., Smith, T. L ., & Weiner, H. (1987). Life events, depressive symptoms, and immune function. Am J Psychiatry. 144(4), 437-41. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 172 Irw in, M ., Daniels, M ., Bloom, E. T., & Weiner, H. (1986). Life events, depression, and natural killer cell activity. Psvchopharmacol Bull. 22(4), 1093-6. lessor, R. (1982a). Critical issues in research in adolescent health promotion. In T. J. Coates, A. C. Petersen, & C. L. Perry (Eds.), Promoting Adolescent Health (pp. 447-465). New York: Academic Press. lessor, R. (1982b). Problem behavior and developmental transition in adolescence. Journal of School H ealth. 52(5). 295-300. lessor, R., & lessor, S. L. (1977). Problem behavior and psychosocial development: a longitudinal studv of youth. New York: Academic Press. Johnson, J. H. (1986). In Life events as stressors in childhood and adolescence Newbury Park, CA: Sage. Johnson, J. H ., & M cCutcheon, S. M . (1980). Assessing life stress in older children and adolescents: preliminary findings with the Life Events Checklist. In I. G. Sarason & C. D. Spielbeiger (Eds.), Stress and anxiety Washington, DC: Hemisphere. Joreskog, K. G., & Sorbom, D. (1988). LISREL VII: A Guide to the Program and Applications. Chicago, IL: SPSS Inc. Jorgensen, R. S., & Dusek, J. B. (1990). Adolescent adjustment and coping strategies. Journal of Personalitv. ^ ( 3 ) , 503-513. Judd, C. M ., & Kenny, D. A. (1981). Process analysis: Estimating mediation in treatment evaluations. Evaluation Review. 5, 602-619. Kanner, A. D., Coyne, J. C., Schaefer, C., & Lazarus, R. S. (1981). Comparison of two modes of stress measurement: Daily hassles and uplifts versus major life events. Journal of Behavioral M edicine. 4(1), 1-39. Kaplan, H. L , & Sadock, B. J. (1991). Synopsis of psvchiatrv : Behavioral sciences, clinical psvchiatrv (6th ed.). Baltimore: Williams & Wilkins. Kellner, R. (1991). Psychosocial stress and disease: A historical sketch. In N. Plotnikoff, A. M uigo, R. Faith, & J. Wybran (Eds.), Stress and Immunity (pp. 61-68). Boca Raton, FL: CRC Press. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 173 Kenny, D. A. (1979). Correlation and causality. New York: Wiley. Kessler, R. C., Foster, C., Joseph, J., Ostrow, D., Wortman, C., Phair, J., & Chmiel, J. (1991). Stressful life events and symptom onset in HIV infection. Am J Psvchiatrv. 148(6), 733-8. Kiecolt-Glaser, J. K ., & Glaser, R. (1992). Psychoneuroimmunology: can psychological interventions modulate immunity? Journal of Consulting and Clinical PsvchologvtAugi. 569-75. Kiecolt-Glaser, J. K ., Glaser, R ., Shuttleworth, E. C., Dyer, C. S., Ogrocki, P., & Speicher, C. E. (1987). Chronic stress and immunity in family caregivers of Alzheimer’s disease victims. Psychosomatic M edicine. 49, 523-535. Kliewer, W., & Sandler, I. N. (1992). Locus of control and self-esteem as moderators of stressor-symptom relations in children and adolescents. Journal of Abnormal Child Psvchologv. 20(4). 393 -412. Krantz, D. S., Contrada, R. I , Hill, D. R ., & Friedler, E. (1988). Environmental stress and biobehavioral antecedents of coronary heart disease. Journal of Consulting and Clinical Psvchologv. 56(3), 333-341. Lazarus, R. S. (1993). Coping theory and research: past, present, and future. Psvchosomatic M edicine. 55(3), 234-47. Lazarus, R. S., Averill, J. R ., & Opton, E. M. (1974). The psychology of coping: Issues of research and assessment. In G. V. Coelho, D. A. Hamburg, & J. E. Adams (Eds.), Coping and adaptation (pp. 249-315). New York: Basic Book Service. Lazarus, R. S., & Delongis, A. (1983). Psychological stress and coping in aging. American Psvchologist. 38, 245-254. Lazarus, R. S., DeLongis, A ., Folkman, S., & Gruen, R. (1985). Stress and adaptational outcomes: The problem of confounded measures. American Psvchologist. 40(7), 770-779. Lazarus, R. S., & Folkman, S. (1984). Stress, appraisal, and coping. New York: Springer Pub. Co. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 174 MacKinnon, D. P., Johnson, C. A ., Pentz, M. A ., Dwyer, J. H ., Hansen, W. B., Flay, B. R ., & Wang, E. Y.-I. (1991). M ediating mechanisms in a school-based drug prevention program: First-year effects of the Midwestern Prevention Project. Health Psvchologv. 10(3), 164-172. M arcia, J. E ., Waterman, A. S., M atteson, D., Archer, S. L ., & Orlofsky, J. (Ed.). (1993). Ego identitv: A handbook for psvchosocial research. New York: Springer-Verlag. M cCubbin, H. I., Needle, R. H , & Wilson, M . (1985). Adolescent health risk behaviors: family stress and adolescent coping as critical factors. Family Relations. 34, 51-62. McEwen, B. S., & M endelson, S. (1993). Effects of stress on the neurochemistry and morphology of the brain: Counterregulation versus damage. In L. Goldbeiger & S. Breznitz (Eds.), Handbook of stress: theoretical and clinical aspects (pp. 101-126). New York: The Free Press. Mechanic, D. (1983). Adolescent health and illness behavior: review of the literature and a new hypothesis for the study of stress. Journal O f Human Stress. 9(2), 4-13. M ish, F. C. (Ed.). (1983). Webster’s ninth new collegiate dictionary. Springfield, MA: Merriam-W ebster Inc. Mokros, H. B., & Poznanski, E. O. (1992). Standardized approaches to clinical assessment of depression. In M . Shafii & S. L. Shafii (Eds.), Clinical guide to depression in children and adolescents (pp. 129-155). Washington, DC: American Psychiatric Press, Inc. Moos, R. H. (1990). Coping responses inventory youth fonn preliminary manual. Palo Alto, CA: Center for Health Care Evaluation, Department of Veterans Affairs and Stanford University M edical Centers. Moos, R. H ., Brennan, P. L ., Fondacaro, M . R ., & M oos, B. S. (1990). Approach and avoidance coping responses among older problem and nonproblem drinkers. Psvchologv And AgingtM arJ. 31-40. Moos, R. H ., Cronkite, R. C., Billings, A. G., & Finney, J. W. (1984). Health and Daily Living Form M anual. Palo Alto, CA: Social Ecology Laboratory, Dept, of Psychiatry and Behavioral Sciences, Stanford University. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 175 M oos, R. H ., & Schaefer, J. A. (1993). Coping resources and processes: Current concepts and measures. In L. Goldbeiger & S. Breznitz (Eds.), Handbook of stress: theoretical and clinical aspects (pp. 234-257). New York: The Free Press. M razek, P. J., & Haggerty, R. J. (Ed.). (1994). Reducing risks for mental disorders: Frontiers for preventive intervention research. Washington, DC: National Academy Press. Newcomb, M . D., & Harlow, L. L. (1986). Life events and substance use among adolescents: M ediating effects of perceived loss of control and meaninglessness in life. Journal of Personalitv and Social Psvchologv. 51(3), 564-577. Newcomb, M . D., Huba, G. J., & Bentler, P. M . (1981). A multidimensional assessment of stressful life events among adolescents: derivation and correlates. Journal of Health and Social Behavior. 2 2 . 400-415. Newcomb, M . D., Huba, G. I , & Bentler, P. M . (1986). Life change events among adolescents: An empirical consideration of some methodological issues. Journal of Nervous and Mental Disease. 174(5), 280-289. O’Dowd, M . A. (1988). Psychoneuroimmunology and H IV infection. A ids. 2(Suppl 1), 201-205. O'Grady, D , & M etz, J. R. (1987). Resilience in children at high risk for psychological disorder. Journal of Pediatric Psychology. 12(1). 3-23. Ouellette, S. C. (1993). Inquiries into hardiness. In L. Goldbeiger & S. Breznitz (Eds.), Handbook of stress: theoretical and clinical aspects (pp. 77-100), New York: The Free Press. Pearlin, L. I., & Schooler, C. (1978). The structure of coping. Journal of Health and Social Behavior. 19(March), 2-21. Pechacek, T R , Murray, D. M ., Luepker, R. V., M ittelm ark, M. B., Johnson, C. A ., & Shutz, J. M . (1984). M easurement of adolescent smoking behavior: rationale and methods. Journal O f Behavioral M edicine. 7(1), 123-40. Pentz, M . A. (1985). Social competence and self-efficacy as determinants of substance use in adolescence. In S. Shiffman & T. A. Wills (Eds.), Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 176 Coping and substance use (pp. 117-142). San Diego, CA: Academic Press. Pentz, M . A. (1994). Prim ary prevention of adolescent drug abuse. In C. B. Fisher & R. M . L em er (Eds.), Applied Developmental Psychology (pp. 435-474). New York: M cGraw-Hill, Inc. Pentz, M . A ., & Chou, C.-P. (1994). M easurement invariance in longitudinal clinical research assuming change from development and intervention. Journal of Consulting and Clinical Psychology. 62(3), 450-62. Pentz, M . A ., Dwyer, J. H ., MacKinnon, D. P., Flay, B. R ., Hansen, W. B., Wang, E. Y., & .Johnson, C. A. (1989a). A multicommunity trial for primary prevention of adolescent drug abuse. Journal of the American M edical Association. 261(22), 3259-3266. Pentz, M . A ., MacKinnon, D. P., Dwyer, J. H ., Wang, E. Y.-I., Hansen, W. B., Flay, B. R ., & Johnson, C. A. (1989b). Longitudinal effects of the M idwestern Prevention Project on regular and experimental smoking in adolescents. Preventive M edicine. 18, 304-321. Pentz, M. A ., MacKinnon, D. P., Flay, B. R ., Hansen, W. B., Johnson, C. A ., & Dwyer, J. H . (1989c). Prim ary prevention of chronic diseases in adolescence: effects of the Midwestern Prevention Project on tobacco use. American Journal of Epidemiology. 130(4). 713-24. Pentz, M . A ., Pentz, C. A ., & MacKinnon, D. P. (1988). Adolescent stress, dmg use, and blood pressure. American Journal of Epidem iology. .128, 891-892. Perry, C. L ., & Jessor, R. (1985). The concept of health promotion and the prevention of adolescent dm g abuse. Health Education Q uarterly. 12(2), 169-84. Perry, C. L ., Stone, E. J., Parcel, G. S., Ellison, R. C., Nader, P. R ., Webber, L. S., & Luepker, R. V. (1990). School based cardiovascular health promotion: the child and adolescent trial for cardiovascular health (CATCH). Joumal of School H ealth. 60(8), 406-413. Piaget, J. (1972). Intellectual evolution from adolescence to adulthood. Human Development. 15. 1-12. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 177 Padioff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psvchological M easurement, i , 385-401. Rahe, R. H. (1974). Life change and subsequent illness reports. In E. K. E. Gunderson & R. H. Rahe (Eds.), Life stress and illness (pp. 58-78). Springfield, IL: Charles C. Thomas. Rahe, R. H ., Pugh, W. M ., Erickson, I , Gunderson, E. K. E ., & Rubin, R. T. (1971). Cluster analysis of life changes, I. Consistency of clusters across large Navy samples. Archives of General Psvchiatrv. 25, 332. Ratliff-Crain, J., Temoshok, L ., Kiecolt-Glaser, J. K ., & Tamarkin, L. (1989). Issues in psychoneuroimmunology research. Health Psvchologv. 8(6), 747-52. Roosa, M . W., Sandler, I. N ., Gehring, M ., Beals, J., & Cappo, L. (1988). The Children of Alcoholics Life-Events Schedule: A stress scale for children of alcohol-abusing parents. Joumal of Studies on A lcohol. 49(5), 422-429. Rutter, M . (1985). Resilience in the face of adversity: Protective factors and resistance to psychiatric disorders. British Joumal of Psvchiatrv. 147. 598-611. Sabshin, M . (1972). Twenty-five years after Men U nder Stress. In D. Offer & D. X. Freedman (Eds.), M odem Psvchiatrv and Clinical Research: Essays in H onor of Rov R. Grinker. Sr. (pp. 94-101). New York: Basic Books, Inc. Sarason, I. G., Johnson, J. H ., & Siegel, J. M. (1978). Assessing the impact of life changes: Development of the Life Experiences Survey. Joumal of Consulting and Clinical Psychology. 46(5), 932-946. SAS Institute Inc. (1988). SAS/STAT User's Guide. Release 6.03 Edition. Cary, NC: SAS Institute Inc. Schaefer, J., & Moos, R. H. (1991). Work stressors and coping among long-tenn care staff. Palo Alto, CA: Center for Health Care Evaluation, Department of Veterans Affairs and Stanford University M edical Centers. Schlesinger, M ., & Yodfat, Y. (1991). The impact of stressful life events on natural killer cells. Stress M edicine. 7(1), 53-60. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 178 Selye, H . (1976). The stress of life (revised ed.). New York: McGraw-Hill Book Co. Selye, H . (1993). History of the stress concept. In L. Goldbeiger & S. Breznitz (Eds.), Handbook of Stress: Theoretical and Clinical Aspects (pp. 7-17). New York: The Free Press. Skinner, H. A ., & Lei, H. (1980). The multidimensional assessment of stressful life events. Joum al of Nervous and M ental D isorders. 168. 535-541. Smith, R. E ., Smoll, F. L ., & Ptacek, J. T. (1990). Conjunctive moderator variables in vulnerability and resiliency research: Life stress, social support and coping skills, and adolescent sports injuries. Joumal of Personalitv and Social Psychology. 58(2), 360-370. Sobel, M . E. (1982). Asymptotic confidence intervals for indirect effects in stmctural equation models. In S. Leinhard (Eds.), Sociological methodologv 1982 (pp. 290-293). San Francisco: Jossey-Bass. Sobel, M . E. (1986). Some new results on indirect effects and their standard errors in covariance stmcture models. In N. Tbma (Eds.), Sociological methodologv 1986 (pp. 159-186). San Francisco: Jossey-Bass. Spielbeiger, C. D., Gorsuch, R. C., & Lushene, R. E. (1970). Manual for the State-Trait Anxietv Inventoiw. Palo Alto, CA: Consulting Psychologists Press. Spirito, A ., Stark, L. J., & Williams, C. (1988). Development of a brief checklist to assess coping in pediatric populations. Joumal of Pediatric Psvchologv. I I , 555-574. Stone, A. A ., & Neale, J. M . (1984). New measure of daily coping; Development and preliminary results. Joumal of Personalitv and Social Psvchologv. 46(4), 892-906. Stone, G. C., Cohen, R , & Ader, N. E. (Ed.). (1979). Health psvchologv: A handbook. Theories, applications, and challenges of a psvchological approach to the health care svstem. San Francisco: Jossey-Bass. Swearingen, E. M ., & Cohen, L. H. (1985). Life events and psychological distress: A prospective study of young adolescents. Developmental Psvchologv. 21(6), 1045-1054. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 179 Tàylor, S. E. (1990). Health psychology: The science and the field. American Psychologist. 45(1), 40-50. Tobler, N. S. (1986). Meta-analysis of 143 adolescent drug prevention programs: quantitative outcome results of program participants compared to a control or comparison group. Joumal of Drug Issues. 18. 535-567. Thshnet, L. (1966). The uses of adversity. New York: Thomas Yoseloff. W idaman, K. F. (1985). Hierarchically nested covariance structure models for m ultitrait-multimethod data. Applied Psvchological M easurem ent. 9(1), 1-26. Wills, T. A. (1985). Stress, coping, and tobacco and alcohol use in early adolescence. In S. Shiffman & T. A. Wills (Eds.), Coping and substance use (pp. 67-94). San Diego, CA: Academic Press. Wills, T. A. (1986). Stress and coping in early adolescence: Relationships to substance use in urban school samples. Health Psychology. 5(6), 503-29. Wills, T. A ., & Shiffman, S. (1985). Coping and substance use: a conceptual framework. In S. Shiffman & T. A. Wills (Eds.), Coping and substance use (pp. 3-24). Orlando, FL: Academic Press, Inc. Wills, T. A ., Vaccaro, D., & M cNam ara, G. (1992). The role of life events, family support, and competence in adolescent substance use: A test of vulnerability and protective factors. American Joum al of Community Psychology. 20(3), 349-374. Windle, M . (1992). A longitudinal study of stress buffering for adolescent problem behaviors. Developmental Psychology. 28(3), 522-530. Wolff, H. G. (1968). Stress and Disease revised and edited bv Stewart W olf and Helen Goodell (2nd ed.). Springfield, IL: Charles C. Thomas. Workman, E. A ., & La Via, M . F. (1991). Stress and immunity: A behavioral medicine perspective. In N. Plotnikoff, A. M uigo, R. Faith, & J. Wybran (Eds.), Stress and Immunity (pp. 69-80). Boca Raton, FL: CRC Press. Zautra, A. J., & Wrabetz, A. B. (1991). Coping success and its relationship to psychological distress for older adults. Joum al of Personality and Social Psvchologv. 61(5), 801-810. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. I A p p e n d i x University of Southern California Health Behavior Survey OFFICE USE ONLY; DO NOT FILL IN THIS SECTION Subject Code Codes 2 3 Survey Date CO 1 8 0 1990eil ny-ni Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 181 ! INSTRUCTIONS 1. Read each question carefully. Marie the box next to youf answer eke this. 2. M you have any questions, raise your hand. 3. PLEASE ALL IN YOUR SEX. BIRTHOATE. GRADE, AND ZIP CODE What Is the zip code 5 ^ girthdate _ _ _ _ _ Grade where you five? M F I I I - I I 1 - 1 I I 6 7 8 9 1 0 1 1 12 I I I I I Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 182 Ouing the next tnhuts. vntts dfovm up 10 4 things thst can happen H Gonteone drinks alcahol(>eer,vrine.ocllquot). Write es many as you can but dont wcfiy K you leave Wank «paces. After one minule oo on to OuesSonZ. 1. X jy/O I-lA t XVot-lT 2. -r n J 0 r -2 .T r ft. T voi- tMr. iv o L ÎÏT 4, WtiaHsvourfa<het*Klob? (ft you dont (ve wfth your father, write your stepfather's or other male guardian's Job.) X t / « 2 .P J 0 ror otto* U c e O n ly 0 1 2 3 e S 6 7 e s What is vour mother's job? (ft you dont Sve with your mother, write your stepmother's or other female guanSan's job.) X«/o?-NTc>_____________ l - o r U f t i o e u » c u n ly 23456769 What sr*ool(s) dkl you attend In the 2lti grads? ■ X s/O it-A Sa______________ _____________ What Is the hiphert grade In school your ÊSSaeom pIsted? p u 0 :} □ 8th grade or less D Some high school □ Graduated from high school □ Vocational or business school □ Some college □ Graduated from college □ Attended graduate or professlona! ' school What is the highest grade in school your n s t o s completed? ^ o£ - M h ^ □ 8th grade or less □ Some high school □ Graduated from high school □ Vocational or business school D Some college □ Graduated from college □ Attended graduate or professlonsi scftool These quesSons are about OGARErtE SMOKING None Parlor allot 1 do. 2-10 C r'dS . 11-20 (Ais 1-5 packs 6-10 pasfg More than :10.C3fkS 7. How many cigaretles have you smoked In vour whole He? 3 .V 0 7 - N U C ft you answered none. SKIP to 12. r O r O >□ . O , o , o tD 8. How many cigaretles Itave you smoked In the oast month (30 days)? rO :0 ,o rO , o . rO tP 8. How many cigaretles have you smoked kl the last week (?days)? r O r O ,o rO jD «D ,o 10. How many cigarettes Itave you smoked in the oast 24 hours? rO :0 r O rO rO .O rO Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 183 11- How many times ^ any) tiave you tried to 13. How many otyrxif do se friends smoke atop Sfflokina? XU 11 - a TC dgarelles? x u / 2 - C - f f - J O 1 never started tfflcMns 1 □ None a □ None 1 □ 1 , D Once s □ 2 4 O Twice O 3 0 (4 3 □ 3 to 5 times a □ 5107 t O 6to 9 tim es < □ 3 to 10 ’ □ to or more times , o More than 10 12. Do you think you wB smoke a doarette In the next two M o lk ? x v t z ^ c c f l , n IdeSnhety wia 1 O 1 probably w0 , D 1 probably wH not □ 1 rJefinitely w D I not More S10 11-20 21-100 lhan Tliese questions are about SMOKELESS TOBACCO. kKiocSng diewtng tobaoco, cnufl. ptug, and 2-4 doping tobacco. Nevcf Once ftnes times times times 100 lime 14. How many times tiave you used smoW ess tobacco In youryAdS.gS?)^^^i^ lO It you answered nevef, SKIP to 17. :0 sD ,D 15. How many times tiave you used smokeless tobacco kl the last montti GO days)? 16. How many times have you used smokeless tobacco In die last week f7 days)? jD >n sO ,o 1 ALCOHOL DRINK - 1 BEER - 1 GLASS OF WINE - 1 MIXED DRINK OF LIQUOR These questions are about drinking ALCOHOL including BEER. WINE, and LIQUOR None 17. How many alcohol drinks have j you evef l>ad in your w h ^ M you answered none. SKIP to 25. Part O f anof 1 drink :0 2 -4 drinks 5-10 drinks 11-20 drinks )0 More 21-100 than drinks 100 drink Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 184 Î9. None How iinny atoohd difnis have y x j had In the fa a jn asei How m aty afcohol drinks have you had In the last week (rday^ 7 :0 Part Of afiof ItW ok :0 :0 2-4 ddnks ,o ,o 5-10 drinks vO 11-20 ddnks ,o Mora 21-100 than drinks 100 drinks .Ü jD " ,o 20. Have you ovof boon drur*7 Xt/24>,pnt< 22. When you drink alcohol (beer, wine, or Squot), how many drinks do you usuaBy I O 1 □ Mo itim e I t / 2 2 - O M > O 2to4fim es ■ □ 1 don*t drink alcohol □ 5 to to times > □ {used to dnnk, but now 1 don't ] □ 11 toaO ftnes > □ Part or aQ of one drink . o 21 to too times □ 2 drinks o Mote tfian ICO times j □ < O 3 drinks 4 drinks 21. kl the last month (30 days}, how many times r □ 5-6 drinks have you been drank? , O 7*6 drinks f □ 9-10 drinks • □ rve never been drunk K □ More than 10 drinks I □ rve not t>een drunk In the last month , O 1 time n 2 (x3tim es : O 4 to 6 tim es □ 7 to 10 times , o More ttian 10 times 23. H you used akxihol during the IssLïSar. how cflen did you use It in each ol the following situalkxis? 3H/23-YWA 0 tf you didn't drink alcohol Not Afewot Some of Most of Every during the last year, check at an tire times the times the time time a this box and SMP to 24. V/hcn you % vere atei>c_ : 0 jO . □ sO b. With just 1 o r2 o th er people^. lO : 0 >□ . a sD c When people over age 30 were present. : 0 ] 0 j D d. A Iap arty _ iD : 0 jD . 0 j O c. I< /2 it> A P Y When your date was present... rO : 0 iD . 0 sO i X1/2ÎEA 0T During the daytime (before AM) p jn .)_ -T,,v >T .livr iD : 0 jD . 0 , 0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 185 g. A ltd u o L - ■X<J-iîCft^C h. m a c a r _ -ujXirihCJX. Not B U b S :0 ■ n Atowof theCmes 2 Û iD Some of Mostof the times the f me >□ >□ <□ .O Every ms sO sD 24. How many tim s s ^ any) have you tried to (top dtinldng sinoitol bevecsges? ZtfyVi-aTm □ I never drink alcohol □ None □ Once O Twice D 3-5 times D G-Sdmes D 10 Of mote times 25. Do you think you wit drink alcotwHn ttie next two 1 □ I definHely will 1 □ } probably wfll 9 □ I probably win not 4 O I definitely will not 26. How many of your dose friends drink alcohol (beer, wtrva or liquor)? 28. When you drink doohol, how often do you drink straight out of the can or bottle rather than from a glass? 1 □ I dont drink alcohol a D I never drink alcohol straight out of thi can or bottle 1 n I hardly ever drink alcohol straight out of the can or bottle 4 □ I sometimes drink alcohd straight oui of the can or bottle 9 □ I fitwgys drink alcohol straight out of tf)8 can or t^ottle 29. These questions relcf to warrwo labels that mav_be written on cans arxf bottles of beer. M ne^nd liquor. Cheefc^ for those statements that ^ tf u n k ^ on the labels, and No for the onesJhaTyou think are not on the labels. ^ fis □ None O 1 fi. Alcohol use impairs the O 2 ability to drive. I □ 2 D □ 3 or 4 O 5 to 7 b. Alcohol use can cause □ 8 to 10 family problems. I □ 1 □ □ More than 10 c. Alcohol use by a Have you seen warning labels on alcohol pregnant woman can harm beverage cans or botbes? the baby. 1 D 2 D D Yes, definitely d. Alcohol use can lead to □ Probably addiction. I O J D □ 1 don't think so O No e. Alcohol use can cause health problems. 1 □ 2 O f. Alcohol use impars the ability to operate machinery. I D 2 D Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 186 UKse quesüoos am aboul marijuana Of POT. None 30. How many (m es have you isad m«*Hna In vour whole life? >□ tf y o u answefed ffone, SWP to 36. 31. How many times have you usod marÿiana In the last month podays)? : 0 22. How many times have you used maiquana In the last week 1 ÊDC 2-4 5-10 Cmgs 11-20 21-100 times. More than lOOftnes : 0 sD 4Ü sD . 0 7 0 jD ,o 33. If you smoke mai^uana, how many 34. How many times (if any) have you tried to JtAils/reefet* do you usuaify tiawo? (If you stop using marguana? smoked with others, count only tfre amount you smoked.) 1 □ rve never used marijuana : □ None , D 1 don’t use marijuana > O Once I □ 1 used to smoke marijuana, txjt now 1 D Twice don’t 3 □ 3-5 times , □ None . O 6-9 times 4 n less than 1 joint 7 □ 10 or more times « O 1 joint 4 □ 2 joints 7 □ 3-4 joints t □ 56 joints 7 O 7-6 joints > e □ 9-tO joints □ More tfian 10 joints . 35. K you used ma/^uana during the last year, how often did you use It In each of the following situations? 0 If you didn’ t use marijuana during tfre last year, check ttiis txix and SKIP to 36. Not at Bit Atewof the times Some of the times Most of ttie time Every time a When you were afore— iD rO sO 4O 3O t> . tW thjustl or 2 other people._. iD j O lO . 0 3O a When people over age 30 were present— lO 7 0 7 0 4O 7O d. At a parly.... 7 0 7 0 . 0 3O When your date was prescnL..., 7 0 7 0 . 0 7 O Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 187 0 - h. Ourfng Itw daytiow (betwe ■4«l p -ra )_ M t c h o a l_ _ fciacar__ N o t B ien iD lO ■D Afewof theftnts :0 iD > □ Som e of the «m es >□ »n ,o Mostof «te «me 40 40 4 0 B » y «me sD jD jD 36. Do you think you via use 37, How many of your dose triends use roatSuana In ttw next two matÿiana? OSOttEt? : O None 1 O 1 deCnitety veil 3 O 1 1 O lprol»l>lywa ] O 2 s O 1 piot>£l]ly wfl not 4 O 3 Of 4 4 O 1 definltoV will not 3 O 5 to 7 4 O 8 to 10 > O More ttian 10 Yes, Id o n t defjottety fVctoablv ttiink so ,'38.'/ Can drinking alcohol during • • pregnancy cause tr ir tf ) defects? :0 , o vO 39. Can drinldng alcohol impair your atjility to work witti macftineiy? iD :0 jD 40 4 0 . Can drinking alcohol impair your ahffity to drive a cai? jD jD 40 4 1. Can drinking alcohol iead to (leallh problems? lO :0 jD 40 42. Can drinking alcohol lead to back problems? lO lO jD 40 These questions are about COCAINE. Including CRACK. None 1 «me 2 -4 «mes 5 -1 0 «mes 1 1 - 2 0 «mes 21-40 «mes More than 40 times 43. How many «mes trave you used cocaine in your wtvole Me? ,o :0 :0 40 ,o 40 lO It you arrswered none. SKIP to 47. 44. How many times have you used cocaine kl the oast month (30 days)? ,o jD iD vD 3 0 ,D ,o Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 188 as. tfthal methods liave you used for taidng cocdne? (t%ock A ll that apply.) 47. How many Of your close tdands use cocaine? , O 1 Itave never used cocaine □ None 1 O Snifling or 'snorting* O 1 1 O Smoking □ 2 4 O ktjection □ 3 or 4 j C Inhaling fumes O 5 to ? 4 O By mouth □ 8 to 10 7 O Other Cl More ttian 10 46. How many times (if any), have you tried to slop using cocaine? 1 O rve rtever used cocaine 48. Do you think you win use cocaine In the next two months? 7 O None 7 □ 1 definitely win 1 O ^nce 1 □ I probably will 4 O Twice I □ 1 probably rw ffl not I O 3-5 b'mes 4 0 6-9 times 7 O 10 or more limes □ 1 definitely will not f iow many times have you used each ct the tollovring dnigs In your '"hole Me? None 1 fm e Heroin, morphine or opium (other than prescribed by adodor) ISO or add D ow ners'Re sleeping pins, bartriturates. tranquilizers or Quaaludes (other than prescnbed by a doctor) Uppers" Cke speed, amphetamines or Ice (other than prescribed by a rtoctor] Oefiberate snifling o< glue, paint, gasoline, or ottier Inhalants to get high « you answered NONE to A a of these. Stqp to 52. ,o ,0 rO rO rO rO 2-4 times sO >□ iD ,o S-10 4Ü 40 1t-20 times jO rO >D rO 21-40 times 4 0 4 0 IMore than 40 times rO rO rO rO Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 189 sa b. c. 51. When was the œsajSSSOLliaifi you used each o( the foOowing dnigs? Heroin, nxxphine or opium (other than prescribed by ■ dodoi) ISO or ad d *Oowneis'Bee sleeping pass, barbiturates, tranqdlizefs or Quaaludes (other than prescribed by adoctoi) Uppers'Ske speed, amphetamines or ice (other than prescrgrad by a doctor) OeBrerate sniffing of glue, paint, gasoEne, or other Intialants to get high Mote than i«s» Never 1 year boo ygac :0 rO > □ > a .D jD ,o Last month 40 40 Last week ,o 5 0 5 0 How many times (K any), have you tried to slop using dnrgs (other than tobacco, alcohol, marÿuana, or cocaine)? O I never started using drugs D None O Once O Twice O 3-Slimes O 6-9 times D to or more Umes 52. b. c. d. For trow long have you used eadi of ttre followirrg drugs gt Cigaretles Smokeless totracco lO Atcotid lO Mar^uana lO Cocatna or crack lO Heroin, morphine, or opkrm iD ISO or add lO •Downers* like sleeping pills, bartriturates, tranquilizers or Ouaalurtes rO Uppers' like speed. Less ttian 1 month rO rO rO rO rO rO rO t-t2 montfis iD rO rO :0 rO rO iD More ttian 12 months 40 4 0 4 0 4 0 4 0 4 0 40 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 190 Yes, I dont deW ety Protably OTnfcso K a To die best o( your knowledge, is Oieca a law mquMng wamkig labels on cans and botdes of beer, wine, and Sqixx? : 0 To the best of your knowledge, is there a law cequifing warning labels on packages of cigarettes? lO SS. Have you ever reoeived any kind otpniiessional counseling, treatment, or therapy tiecauseot your use No. never Yes, but not kl the past 12 months Yes, sometime In the past 12 montfis a Tobacco iD : 0 >□ b. Alcohol >□ : 0 , o t Marÿuana lO jD lO d Cocaine (or crack) , o lO lO a Any odier drug iD iD j O 56. Have you ever been admitted to an emergency room or clinic because of your use of: a Tobacco b. Alcohol c. Marÿuana d. Cocaine (or crack) « ■ Any other drug No. never .□ ■ D .D .□ .O Ves,t)ut not in the past 12 months :D rO rO :0 :0 Yes, sometime In ttie past 12 montfis iD lO lO lO St. Has anyone in your family ever received any kind of professional counseling, treatment, or ttierapy because of their use o t No. never Yes. but not ki the past 12 months Yes. sometime hi the past 1 2 months a Tobacco ■ O ;0 iD b. Alcotioi iD rO , o Mar^uana iD lO , o id. Cocaine (or crack) iD jD jD Any other drug iD , n Reproduceid with permission of the copyright owner. Further reproduction prohibited without permission. 191 h H s a Whal Qiades do you usually get in school 61. How many hours a day do you spend watching TV (on an average weekday)? □ Mott/yA'e D A's and B's □ 8 or more hours a day □ Mostly B's □ 5 to 7 hours □ B's and C s D 3 o re hours D Mostly C s O 2 hours O C s and O's D 1 hour D Mostly O's O None □ D 's a n d F s D Mostly F s 62. Hrrw ofien do you attend church or tempte? 59. How many schools have you gone to since , D Often (about once a week) irct grade? 3 O Sometimes 3 O Hardly ever □ 1-2 4 O Never o % O 5 ^ 63. Do you Sve with: D 7-8 D 9-10 O Both parents O More than 10 O Mother only O Mother and stepfather 60. How many people (counting yoursetl) Sve in D Father only your house? □ Father and stepmother □ Other relatives O 2 □ Guardian(s) □ 3 o r4 D Boyfriend/girtfriend □ SorG □ Other O 7 o r8 □ 9 or more 64. Out ol every 100 people your age, how many do you think use each of the (oRowino drm s at least once a month? C l a £Q W 52 5Q 63 ZQ ËQ 2Q JSQ a. Cigarettes iD : 0 jD .O sO rD lO ftD vQ teD 310 b. Alcohol rO : 0 >□ , 0 3O 4O 7O iQ f D 30O 310 c. Marijuana : 0 : 0 j D 3 0 sQ 3O rO 3O fO 3*0 3iD ' d. Cocaine (or crack) iD dO , 0 rO 3O «Q 7O aO 7O 3*0 3 3 0 «. Heroin : 0 sO 3 0 3O * 0 iD aO * 0 3*0 310 t U ppers' (amphetamines or i» ) rO j D >0 3 0 3O aO 7O aO #D 3*0 u D g "Downers'(bartdurales) .□ : 0 3 0 3 0 3O aO 7Q # 0 * 0 3*0 33O h. Inhalants felue, paSit, gasoline) iD : 0 j O 3 0 *D *D tD * 0 * 0 3*0 33D L Drive under the influence of drugs or alcohol iD iD jD 3 0 3O aO 7O aO * 0 3*0 3 3 0 Reproduced with permission of the copyright ow ner Further reproduction prohibited without permission. 192 £5. 0 your best fiondo(féfed)rou the loQowfng drugs. t)ow hard %vouldtt be to feAisettteof^er? Very Vety hard H ad East easy m . Oga><it<es •D : 0 3 0 vO b. AJootio( iQ j D 3 0 . C Matïuana •n : 0 3 0 4O d. Cocaiiw (orcradi) iD 3 0 3 0 . 0 Heroin lO 3 0 3 0 . 0 t ■Uppeo* (amphetamines or ioe) : 0 3 0 3 0 rO g. TJownere* CarWuraJes) , o 3 0 sO , o h. Inhalants blue, paM, sasoCne) lU 3 0 3 0 . 0 es. How much do you think toeing people using drogs on TV, in magazines. Of In movies can influence people your age to use one o( these drugs? A None oreatdea! Some AHtle at all Tobacco .D 3 0 3 0 "O. b. Alcohol , o 3 0 3 0 3O c Marÿuana : 0 3 0 3 0 vO d. Cocaine (Of crack) lO 3 0 3 0 iO e. Olhef drags rO 3 0 3 0 , 0 67. K vou have used anv ot the foflcwino draos in the last year. did you use them with any other drug at tt>e tame time, or rigid after one another? (Ctieck all drugs used at the same time.) D .ft you dklnT use any drags Other in the last year, check fits. Mac Css. "Uooers" "Oownets" Heroin .Drug this box end SKIP to 68. a. __ O O O a 0 D 0 b. Alcohol with....... ................. . o o a 0 0 0 c. ktarÿiana with .......... . o a 0 0 0 d. Cocaine with.. _ _ a 0 0 0 Uppers ’with .... 0 0 □ (. "Downers' with........ __ _ 0 D g Heminwith........... 0 68. ft you wanted to use each ot the following drugs, Itow easy would it t>e tor you to get some? Very Very hard Hard Easy easy Cgarettes iD 3 0 3 0 4O b. Alcotwl iD 3 0 j O .O Mat^uarta .□ 3 0 3O 3O d. Cocaine (or crack) tO 3 0 3O 3O Heroin 3 0 > a 4 0 1 "Uppers" (amphetamines or kx) lO 3 0 3O 4O g "Downers" (barbiturates) lO 3 0 )D . 0 h. Intralants (glue, paint, gasoline) jD 30 . n . 0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 193 . es. H you haw used any ofttte following dnigs, check the MOST IMPORTANT REASONS for your taking I ' aachdnjQ. (Check ALL that aoptyj M vou haw nol used anv ot these dnxg-SM P to 71. I Crfgsmtl^ Alcohol Cocaine Marijuana torgackl Other lèiias a. To e»penmonl-fo see what Ifs ike . a rO aO aO aO b. To relax : 0 : 0 aO a d aO c. To foel good or get high : 0 aO aO aO < L To haw a good time with my friends : 0 aO aO aO c To Tel go" wM h a date iD aO aO aO t To get away from my problems or troubles iD >□ >□ aO aO g- Because ol boredom, nothing d se to do tO aO >□ aO aO h. Because of pressure from others : 0 aO .O aO aO L Because of anger or fnistmtion aO aO aO aO t To get through the day iD aO >□ aO aO 70. U you haw ever oiê! using any of the following drugs, check the MOST IMPORTANT REASONS for qaSrtg. {Chock £LL fbst appfy.) Kyoij have not quS using any of these drugs, SKjP to 71. Cocaine other Cioarettes Alcohol Marituana forcradd gass a. Too expensive lO x d x d x d i d b. Pressure from others ■ D i d i d x d i d c . fell bad about myself for using iD x d i d x d i d d. It was altocting my school work ,o x d i d x d i d e. Had a bad reaction and/or H made me skrkiu i d i d x d i d t It was affecting my relationships with others 1 □ x d i d x d i d a - My friends ryiN using ,o x d i d x d i d h . Got new friends that rSdnT use , o x d i d x d i d L Just deckfed 1 dkfnT want to do It i d i d i d x d i d 71. How many times in the last year have you tried to talk with any of tfie foOowinr; people ahrxjt rlnig prevention? More than None Once 2-4 times 4 times a Parents : 0 xO i d xO b . Close trierxfs : 0 : d i d x d c . Teacticfs lO x d i d x d d. Others iD x d i d x d 72. Of the two adults wtro are the most important in your (fe, how many do yrxj think use each of ttie following drugs? None One Two a. Cigarettes i d i d i d b. Alcotioi i d i d I d c. Manjuana i d i d i d rf. Crxraine (or crack) i d i d i d e. Otfier rJrugs i d i d i d Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 194 73. Of tie 1*0 aduNs who are (he m ost ImpoflanI In your Hie. h a : «Nhef e w goBsn W o Im ubk (or dm km g atoohot or uskig druQs? , Q None I n 1 , 0 2 74. How much d o you care whether the following people use drugs or alcohol? Not A Very Btall aas Somewtiat much a. Patents ,o ,o b. Close friends ,o ,o ,o c Teachers ,o , D »Q . O d. Offiers ,o , D , □ ,o 75. How much d o you care how each of ttie following would act towards you If you u; Not A Very at alt Effls Somewhat m y ç h a Parents iD , D ,D b. taose friends ,D ,D 4 0 c Teacfiets , □ ,o , □ <Q d. Otfiers iD j D , o < o 76. How would your frierxls act toward you if you used drugs Of drank alcohol? I O Vety friendly 1 D Pretty friendly I O Pretty unfriendly < D Very unfriendly 77. How would your friends act toward you it you smoked cigarettes? ■ □ Very friendly : O Pretty friendly 1 O Pret^ unfriendly ' O Very unfriendly ?8. How important is It for you to get good grades? ■ n Very important I O Important > D A little important ‘ O Not at all important 79. a. b. c. d. f. 9 h. Below is a Est of evente tfiat somefrmes flapped to people your ag e. For each of the events check If tfie event liappened to you in the last 12 MONTHS. m. n. Moved to new fxjme Changed to new scfiool Cfianged to new job Parents divoroed or separated Parents away from liome m ore often Failed a grade" lo s t a d o s e friend Increase in number of argum ents with parents Major ainess o r injury to yourself Broke up with txiyfriend/girifriend Ran away from/leR fiome Gained or lost a lot of weight Quit school Qu it job ÏÊ5 i D i D iD ,o lO iQ lO iD iD .D H a jD j D :D iD , D , D ,o ,n .□ ,o :0 ,o ;0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 195 N on» 80. In the D3!rt year, how many times hav e you been suspended ■ to m school or job for cmoMrtg, drinldng, or using drugs? lO 81. In the oast year, how many fimes have you been ki trouble vdth the pofice for drinldng or using drugs? iD 82. In the last rnpnth (30 davsi. how many times did you drive aRer drinldng alcohol or using drugs? 83. In the last month (30 days), how many times were you with Eomeorw who w as drrvirrg aSer drinldng alcohol o r using dnrgs? lO 84. In the last month (30 days), ixrw many times did you choose rwt to tide wRh srxneorre wtio was driving alter drinking alcohol or using drugs? iD 85. In the lastrnonth (30 days), how many times did you ggt use alcohol o r other dnrgs wtren you had ttie chance? :D 86. In the last month (30 days), trow many times did you try to stop someone from drivirrg alter drinking alcotioi or using drugs? rO 1 D m » :D rO rO 3 o r 4 M ore ttian Cmes 4 tim es >□ >G rO rO ,o jO rO ,0 sO Not Stall 87. How sure are you that you can turn down a ctiance to use alcotioi or other dnrgs in the next two months? lO 88. How much would you care if a friend drove after drinking alcohol or using dnrgs? i O Pttie Romewhat Very ,o xO Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 196 g The asU im c thaï you drove altef drinking a k x M or using dnigs, WHICH DRUGS did you u se? (Check ALLthal vou used.) O I have not driven under the influence ot dnrgs D Alcohol □ Manjuana □ Cocaine □ Other drug SO . During an average week, how rrruch money do you receive from working? D Less than Î1 0 per week □ $10-25 per week O $26-50 per week D $51-75 per week □ $76-100 per week D More than $100 per week 31. During an average week, how much morrey do you receive from other sources (allowances, etc.)? □ Less lhan $10 per week □ $10-25 per week □ $26-50 per week D $51-75 per week O $76-100 per week D More lhan $100 per week 52. On the average, how often do you go out 95. How m arg of the foliowing dnrg education with a date? activities ftave you had in sctvoot? (Ctieck ALL that apply.) I □ Never O Once a month or less r O Special lesson In driver's education : □ 2 or 3 times a month class 4 □ Once a week 1 □ Special course about drugs ; D 2 or 3 times a week j O Films, lectures, or discarssions in one of • □ More than 3 times a week my regular cotrrses < O Student assistance program or peer 53. Have you ever had any dnrg education counseling on drugs courses or lectures in school? 5 □ Films or lectures, outside of my regular courses I D No < D Special discussions about drugs O No, and 1 wish 1 had 1 □ Other ’ D Yes 96. Have you ever particapated in any of the 91. Would you say that the information about following l-STAR activities? (Chectc ALL that dnrgs that you received in srsttool classes or apply.) programs h a s_ 1 O Classroom lessons ' n Made you less interested in trying drugs 3 O Special program in sporls/phys. ed. □ Not changed your interest in trying 5 D Student assistance program dnrgs « □ Homework with parents ' O Made you more interested in trying 3 O Special parent night or parent program drugs i O Rap or video contest ‘ O 1 never received any information abocrt 7 O Rally drugs in school » □ Survey 7 O Community meetings or special events 1 0 □ High school press conference 1 1 n Other Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 197 97. In tha p ast year, how m any times have you skipped going to e ach o f th e following places, ggt because o fB ness? None 3 o r 4 O nce 2 times tim es S o r e tim es More ttian 6 tim es a. b. c. School Wock M eetings/Clubs iD lO lO : 0 > □ : 0 jD , 0 vO 4 0 4 0 s G j D sG 4 0 4 0 4 0 98. Below is a Est o t problems that people sometimes tiave o r feel. For e ac h one. please check HOW OFTEN each problem tiappens to you. Almost Never SomeUmes Often Almost Alwavs a. 1 feel Eke 1 d o n t want to doanytfiing . □ : 0 >□ 4G f a . 1 find It fiard to keep my mind on som ething I'm woridng on lO lO j O 4 0 c. 1 am no good for anything a t all j O 4 0 d. 1 am tired during the day j G 4 0 e. Ifeel lonely iD :0 j G 4 0 1. 1 feel nervous o r anxious iD : □ j G 4 0 g- 1 feel dizzy or Eg titheaded >□ lO i G 4 0 h. 1 g et tieadaches iD iG 4 0 L 1 w ony about things in my life . □ : 0 iG 4 0 i- 1 feel sad lO iG 4 0 k. 1 tiave pains in m y tieart or ctiest :D j G 4 0 I 1 fe d Eke 1 tiave nothing to look forward to :D iG 4 0 m. 1 have fjeen seriously thinking about a way to tuirt myself . □ : 0 .G 4 0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 198 S3. Below is a 6st o f th in g s th a t peo p le d o w h en they have a problem. For e a c h o n e. c h e c k HOW OFTEN you respond o r a c t th a t vray. Almost Almost .Never Sometimes Often Atwav a. 1 get Wormalion that is needed to deal with the problem iD j D j D ^O b. 1 te: myself ttie problem is not worth getting upset atxjut ] □ :D rO 4 0 c 1 just wail and hope ttiat things win get t>etter with time lO : 0 >□ 4 0 d. 1 work it off tiy physical exercise rO rO ,n 4 0 e. 1 go out and play sports rO : 0 3 0 4 0 t 1 talk with my m other or lather rO rO 3 0 4 0 9 1 talk with one of my (rierxls iQ :D 3 0 4 0 h . 1 get mad at people ,o : 0 3 0 4 0 L 1 Esten to music rO : 0 3 0 4 0 i Isay a prayer ,o 3 0 4 0 k. 1 read books o r m agazines, or watch TV rO jO 3 0 4 0 I 1 do something t>ad or cause trouijle rO jD 3 0 .o m. 1 sleep more ,o rO 3 0 4 0 n. f tiy not to think atjout K lO rO 3 0 4 0 0. 1 get away from people iQ rO 3 0 4 0 p. 1 take pills (like aspirin or tranquilizers] to feel txztter ,o : 0 3 0 4 0 q 1 smoke a cigarette to relax rO : 0 3 0 4 0 f. 1 drink alcofx)! or u se dnrgs fo feel belter iD rO 3 0 4 0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 199 Almost Never Somefimes Often Almost Alwavs s. 1 hang out «1th friends • iD >□ vO t ( g o to or have a party >□ : 0 >□ u. 1 try to calm myself lO : 0 rO V . 1 try deep breathing iD : 0 , o 100. Bead each statem ent and ttien check the answer tfrat t>est describes vour SCHOOL None of the fm e Som e of tfie fm e M ostof the fm e A H of the fm e a. Students a te very interested in getting to know other students : 0 rO 4 0 b. Students enjoy helping each otfier with homework rO :D rO .O c. Teachers take a personal interest in students : 0 , o 4 0 d. If students want to talk about something teachers wSi find time to do it iD : 0 iD 4O 101. each statem ent, check the answer that best describes vour FAMILY. None of tfie time Som e of tfie fm e Most of the fm e All of the fm e a My family m em bers realty help arid support one anotfier lO iO )D 4O b. My family m em bers keep tfieir feelings to tfiemsefves rO iQ 4O c. We fight a lot in our family lO ) 0 iO 4O d. There b a feeling of togetfiem ess in our family rO : 0 , o . 0 e. We ten each otfier atrout our personal problems lO rO , o 4 0 Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 200 None of Some of M ostof AH of I the time the time the time the time i 1 My famBy m em bers 1 dlticlzs each othor iD :D iO vD i f i . Someone usuaX ygekupselN -- you c o m p ly k i our family ,o jD »□ vO h. My famBy m em bers hit each other iD :0 >o vO My family m em bers strongly encourage each other to stand up for their lights iD :0 iO vO 1 j We realty get along w dl with 1 each other jD lO .□ k. There are open discussions in our family ,o :D iO vO L 1 am a iKitheaded peisrxi ,o rO ,D vO m . W tw il really g et m ad. Isay nasty things iD : 0 : 0 vO 102. In ttie last how m uch did you u a s J tieip with a problem from each o! the toBowing? ! Not A Very D oes not ela n fiWe Somewhat much a o o lv to m e a. Parent rO iO vO sD b. Close friend iD ;D . j D rO c Teacher lO rO , o vO ,o d. Other iD rO iD vO j O 103. In the last YEAR, how sabsKed were vou with tfie advice given to you by each nf the fnllnwing? Not A Vety Does not at all Bttte Somewhat much aoolv fo m e a Parent :0 :0 - □ sD b Close friend rO >□ vD c. Teactier iD rO :0 vO d. Oilier lO :0 j O vO 1Ü Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. 201 104. Suppose a friend S tarted to have problems tiecause h e (or she) w as drinking a lot of alcohol o r u d n g drugs. Could vou helo that person stop? 105. Suppose a friend started to have problems because tie (or she) w as drinking a lot of alcohol or using drogs. you actually do things to help tfiat person stop? I □ 1 definitely could eg) help , O 1 would definitely d c things to help 2 □ 1 probably could not help 2 O 1 protaably would do things to help - a □ 1 probably could help a □ 1 probably would qqI do things to fielp 4 □ 1 could definitely help □ 1 definitely would c s! d o things to help W as not Jn 1-2 situation None 106. During the o ast year, how m any tim es have you Estened to a triend talk about t h ^ problems with drinking alcohol or using dnjgs? i D : 0 sD 107. During the o ast year, how many tim es have you tried to stop a friend from driving a car alter they had been drinking alcohol or using drugs? l O iD jD 108. During the o ast year, fiow many tim es have you tried to stop a friend from getting in a car being driven try som eone wtro had treen dririkitrg alcohol or using dnrgs? l O . rO rO 109. During the oast 30 days, how m any times tias a d o se friepd offered you alcohol o r otfier drugs? > □ rD rO 3-1 times 5 o r more times rO .a < o A fl of them Most of tfiem Some of them None of them 110. How many of your d o sest friends think it is OK to get dnm k or high every now and tfien? 1 1 1 . How m any of your d o sest friends have told you that it is stupid to drive after having a little alcohol or drugs? r D rO Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Adolescent cigarette use and psychological distress: relation to adult obesity risk

PDF

Emotional intelligence and smoking risk factors in early adolescents

PDF

Aggression and victimization as predictors of drug use in early adolescence

PDF

Effects of parent stress on weight-related parenting practices and child obesity risk

PDF

Motivation and the meanings of health behavior as factors associated with eating behavior in Latino youth

PDF

An investigation of caregiver sense of control and coping strategies in moderating the effects of caregiving demands

PDF

Adaptation to a life -threatening diagnosis: Dispositional optimism and pessimism and posttraumatic growth among patients with HIV

PDF

Social selection and influence of alcohol & marijuana implicit cognitions and behaviors: a longitudinal investigation of peer social network dynamics

PDF

Psychosocial and behavioral ractors associated with emotional eating in adolescents

PDF

Correlates of smoking behaviour among Egyptian adolescents

PDF

Competing risks: the role of the perceived consequences of refusing to share injection equipment among injection drug users

PDF

Intrapersonal and environmental factors associated with Chinese youth alcohol use experimentation and binge drinking behaviors

PDF

A network analysis of online and offline social influence processes in relation to adolescent smoking and alcohol use

PDF

Exploring the role of peer influence, linguistic acculturation, and social networks in substance use

PDF

Cultural risk and protective factors for tobacco use behaviors and depressive symptoms among American Indian adolescents in California

PDF

The acute relationship between affective states and physiological stress response, and the moderating role of moderate-to-vigorous physical activity

PDF

The relationship between high school anti-smoking policies and adolescent cigarette use

PDF

Predictors of adolescent depressive symptomatology

PDF

Non-steroidal anti-inflammatory drugs, anti-acid drugs and passive smoking as impact factors for esophageal and gastric adenocarcinomas

PDF

Using a structural model of psychopathology to distinguish relations between shared and specific features of psychopathology, smoking, and underlying mechanisms

Asset Metadata

Creator Broder, Benjamin Isaac (author)

Core Title Adolescent stress and styles of coping: Predictors and moderators of psychological distress

Degree Doctor of Philosophy

Degree Program Preventive Medicine (Health Behavior)

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag health sciences, mental health,OAI-PMH Harvest,psychology, behavioral,psychology, developmental

Language English

Contributor Digitized by ProQuest (provenance)

Advisor Petz, Mary Ann (committee chair), Anderson Johnson, C. (committee member), Chou, Chih-Ping (committee member), Henderson, Victor (committee member), Pike, Malcolm (committee member)

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c17-501285

Unique identifier UC11352666

Identifier 9630742.pdf (filename),usctheses-c17-501285 (legacy record id)

Legacy Identifier 9630742-0.pdf

Dmrecord 501285

Document Type Dissertation

Rights Broder, Benjamin Isaac

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the au...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus, Los Angeles, California 90089, USA