Close
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Patterns and risk factors for adolescent smoking progression
(USC Thesis Other)
Patterns and risk factors for adolescent smoking progression
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INFORMATION TO USERS This manuscript has been reproduced from the microfilm master. UMI 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 of computer printer. The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send UMI a complete 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 materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. ProQuest Information and Learning 300 North Zeeb Road, Ann Arbor, M l 48106-1346 USA 800-521-0600 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Reproduced with permission of the copyright owner. Further reproduction prohibited without permission. PATTERNS AND RISK FACTORS FOR ADOLESCENT SMOKING PROGRESSION by Chaoyang Li A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (PREVENTIVE MEDICINE - HEALTH BEHAVIORAL RESEARCH) December 2001 Copyright 2001 Chaoyang Li R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. UM I Number: 3065814 Copyright 2001 by Li, Chaoyang All rights reserved. ___ « > UMI UMI Microform 3065814 Copyright 2002 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.O. Box 1346 Ann Arbor, Ml 48106-1346 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. UNIVERSITY OF SOUTHERN CALIFORNIA The Graduate School University Park LOS ANGELES, CALIFORNIA 90089-1695 This dissertation , w ritten by U nder th e direction o f h is ... D issertation C om m ittee, a n d approved b y a ll its m em bers, has been p resen ted to and accepted b y The Graduate School, in partial fulfillm ent o f Chaoyang Li requirem ents fo r th e degree o f DOCTOR OF PHILOSOPHY 3U i o f Graduate Studies D ate 12- 17-2001 DISSER TA TION COMMITTEE /£ '/± 5 /# / R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. DEDICATION To my parents, who made this life possible, To Qiuping, who makes this life enjoyable. To -Jimmy, who makes this life even more valuable. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. ACKNOWLEDGMENTS in This dissertation signifies the culmination of my four years in Graduate School. It is not only a result of hard work by myself, but also a reflection of the contributions made by the people I have encountered since entering the University of Southern California, particularly those on my dissertation committee. I feel very fortunate to have Professor Michael D. Newcomb on the committee. His profound knowledge and productive work on etiology and consequences of drug abuse and expertise of structural equation modeling and quantitative psychopathology have been invaluable to me. I am also very indebted to Professor C. Anderson Johnson for his important advice on smoking epidemiology and prevention strategies across culture and lifespan. I greatly thank Professor James H. Dwyer for his broad exper tise of statistical methods for longitudinal data and his continuous guid ance to me over years. I especially appreciate Dr. Chih-Ping Chou, Asso ciate Professor of Research, for his unique knowledge and expertise of structural equation modeling, latent growth curve modeling, and multi level models and his hand-to-hand instructions to me in the five years. Dr. Chou, together with my best friends, Dr. Li Li and Dr. Guoping Feng, introduced me to the field of health behavioral research five years ago. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. My greatest gratitude goes to my advisor. Professor Mary Ann Pentz. Professor Pentz provided me a great opportunity to enter into the field of health behavioral research five years ago. Over the years, her dedication to research and her philosophy of life have encouraged me to overcome many difficulties in my studies and life. Her insightful exper tise of theories, methodologies, and practice in drug abuse prevention, clinical psychology, mental health, and health behavioral research has been of great importance to me. Particularly during the process of con ducting this dissertation research. I am very grateful to her for her guid ance. patience, encouragement, and most importantly, her belief in my ability to do it. Of course, without the support and permission of Professor Mary Ann Pentz, the principal investigator of the Tobacco Program and Policy Trial (Project TOPP). and the support from Dr. Clyde A. Pentz. the co principal investigator of the Project TOPP, this dissertation would have been impossible. Finally, I sincerely thank the Tobacco-Related Disease Research Program at the University of California for granting me a dissertation research award (9DT-0199). which has provided me a valuable financial support to timely accomplish this dissertation. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. TABLE OF CONTENTS V DEDICATION.....................................................................................................................ii ACKNOWLEDGMENTS...............................................................................................iii LIST OF TABLES..........................................................................................................viii LIST OF FIGURES...........................................................................................................x ABSTRACT........................................................................................................................xi 1 INTRODUCTION....................................................................................................... 1 2 ADOLESCENT SMOKING PROGRESSIONS: A REVIEW......................12 2.1 A n a l y t i c M e t h o d s ......................................................................................................................................14 2 .2 S m o k i n g P a t t e r n s .......................................................................................................................................2 0 2 .3 R isk F a c t o r s .......................................................................................................................................................23 2 .3 .1 Intrapersonal or psycho-behavioral fa c to r s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23 2.3.2 Interpersonal or social fa c to r s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25 2.3.3 Dem ographic F a c to rs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 2 . 4 T h e o r e t i c a l a n d P r a c t i c a l Is s u e s ..........................................................................................33 2 .5 E v a l u a t i o n ok P r e v e n t i o n E f f e c t s ........................................................................................ 3 4 2 . 6 M e t h o d o l o g i c a l C h a l l e n g e s .......................................................................................................3 6 2 .7 F u t u r e D irec t i o n s f o r R e s e a r c h o n a d o l e s c e n t s m o k i n g ..........................3 7 3 STUDY DESIGN AND MEASUREMENTS.................................................... 40 3.1 O v e r v i e w o e S t u d y D e s i g n ...............................................................................................................4 0 3 .2 T h e o r e t i c a l M o d e i .....................................................................................................................................41 3 .3 H y p o t h e s e s ..........................................................................................................................................................4 3 3 . 4 T h e P r o j e c t T O P P ........................................................................................................................................4 4 3.4.1 Intervention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.4.2 P rocedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .46 3.4.3 Data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3 .5 S u b j e c t s .................................................................................................................................................................4 8 3 .6 M e a s u r e s ..............................................................................................................................................................4 9 3 .6 .1 Outcome m easures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 3.6.2 Predictors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3.6.3 Reliability estim ates o f the m easured variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. VI 4 NEW STATISTICAL METHODS FOR ASSESSING CHANGE..............56 4.1 L a t e n t T r a n s i t i o n A n a l y s i s ( L T A ) ..........................................................................................5 9 4 .1.1 O venicw o f sm oking transition m o d els . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 4.1.2 Param eter estimates in L TA m o d e ls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4.1.3 M odel fit index in LTA m o d els . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63 4.1.4 Significance tests fo r param eter estim ates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 4 . 2 L a t e n t G r o w t h M i x t u r e M o d e l i n g ( L G M ) ..................................................................6 6 4 .2 .1 O v e n iew o f Latent Growth M ixture M odeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4.2.2 M odel selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 4.2.3 A lode ling strategy •. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 4 . 3 M e d i a t o r a n d M o d e r a t o r M o d e l s .........................................................................................71 4 . 4 M e d i a t i n g P r o c e s s o r La t e n t G r o w t h C l r y e M o d e l i n g ...............................7 4 4 . 5 In t e r a c t i o n b e t w e e n L a t e n t G r o w t h F a c t o r s .........................................................7 6 4 . 6 M o d e l fit i n d i c e s fo r l a t e n t g r o w t h c u r v e m o d e l i n g ....................................7S 4 . 7 G e n e r a l i z e d E s t i m a t i n g E q u a t i o n s ( G E E ) M o d e l i n g .......................................79 4. 7.1 Overview o f generalized estim ating e q u a tio n s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 4. 7.2 Steps o f GEE m odeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 4 . 8 In t r a c l a s s C o r r e l a t i o n ( I C C ) a n d D e s ig n E f f e c t s .............................................83 4 . 9 S t a t i s t i c a l P a c k a g e s ..............................................................................................................................8 5 5 RESULTS................................................................................................................ 86 5.1 G e n e r a l C h a r a c t e r i s t i c s o f A d o l e s c e n t S m o k i n g .............................................8 6 5 .1. 1 Attrition Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 5.1.2 Prevalence rates o f sm o k in g . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 5.1.3 Comparison o f sm oking prevalence with California, national, and Florida samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91 5.1.4 Changc o f smoking sta g e s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 5.1.5 Patterns o f regular addictive sm oking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 5.1.6 Intraclass correlation (ICC) and design effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 5 .2 L a t e n t T r a n s i t i o n A n a l y s e s o f S m o k i n g S t a g e s .................................................9 6 5.2.1 Cross-validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 5.2.2 M easurement param eters (p p a ra m eters) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 5.2.3 Unconditional probabilities o f latent class m em b ersh ip . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 5.2.4 Transition probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 5.2.5 Omnibus tests o f transitional probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 5.2.6 Comparison o f sm oking stages by m odel and definition . . . . . . . . . . . . . . . . . . . . . . . I 09 5.3 G r o w t h M i x t u r e M o d e l i n g o f S m o k i n g T r a j e c t o r i e s ...............................11 0 5 . 3 .1 Pattern o f smoking trajectories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . I l l 5.3.2 Predictors fo r differentiating sm oking tra jecto ries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 5.3.3 Comparison between SA S PRO C TRAJ and M p lu s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 5 . 4 M e d i a t i n g P r o c e s s o n S m o k i n g T r a j e c t o r i e s ........................................................1 2 4 5 .5 In t e r a c t i o n b e t w e e n F r i e n d s ’ S m o k i n g a n d R e f u s a l S e l f - e f f i c a c y o n S m o k i n g T r a j e c t o r i e s ..............................................................................................................12 7 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Vll 5 .6 G e n e r a l i z e d E s t i m a t i n g E q u a t i o n s ( G E E ) M o d e l i n g o e R e g u l a r a n d A d d i c t i v e S m o k i n g B e h a v i o r s .................................................................................................1 2 9 6 DISCUSSION.........................................................................................................132 6.1 M a j o r F i n d i n g s o f t h e P r e s e n t R e s e a r c h ......................................................................132 6 .2 C o m p a r i s o n o f S m o k i n g P r e v a l e n c e R a t e s w i t h N a t i o n a l , F l o r i d a , a n d C a l i f o r n i a s a m p l e s ...................................................................................................................135 6 .3 S m o k i n g T r a n s i t i o n s d u r i n g E a r l y A d o l e s c e n c e ...............................................138 6 . 4 T r a j e c t o r i e s a n d R isk F a c t o r s o f A d o l e s c e n t S m o k i n g ...........................1 3 9 6 .5 A d o l e s c e n t N i c o t i n e A d d i c t i o n a n d R isk F a c t o r s ............................................142 6 .6 L i m i t a t i o n s a n d S t r e n g t h s ............................................................................................................14 4 6 .7 Im p l i c a t i o n s fo r P r e v e n t i o n , T r e a t m e n t , a n d Po l i c y ....................................14 7 REFERENCES...............................................................................................................150 APPENDIX A ................................................................................................................. 173 P e r c e n t a g e s o f G e n d e r a n d E t h n i c i t y b y S c h o o l D i s t r i c t a n d S c h o o l ....................................................................................................................................................................173 APPENDIX B ................................................................................................................. 174 M e a s u r e s o f t o b a c c o u s e s , s o c i a l i n f l u e n c e , i n t r a p e r s o n a l a n d d e m o g r a p h i c c h a r a c t e r i s t i c s o f si r v e y s a t t h r e e w a v e s ...................I 7 4 APPENDIX C ................................................................................................................. 176 L a t e n t T r a n s i t i o n A n a l y s i s P r o g r a m a n d P a r a m e t e r E s t i m a t e s 176 APPENDIX D ................................................................................................................. 197 S A S C O D E S ...................................................................................................................................................... 197 APPENDIX E ................................................................................................................. 198 M PLU S SCRIPT'S............................................................................................................................................ 19 8 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. LIST OF TABLES T a b l e 2. l T a b l e 2.2 T a b l e 3.1 T a b l e 5. l T a b l e 5.2 T a b l e 5.3 T a b l e 5.1 T a b l e 5.5 T a b l e 5.g T a b l e 5.7 T a b l e 5.8 T a b l e 5.9 T a b l e 5 .10 C a t e g o r i e s o f s t u d i e s o x s m o k i n g p r o g r e s s i o n .............................1 3 S u m m a r y o f s t u d i e s o n s m o k i n g s t a g e s / p r o g r e s s i o n s 1 5 R e l i a b i l i t y e s t i m a t e s o f t h e m e a s u r e d v a r i a b l e s ......................5 5 A t t r i t i o n e l a t e s a t t i m e 2 a n d t i m e 3 b y d e m o g r a p h i c CHARACTERISTICS a.XD SMOKING ST ATI’S AT BASELINE.......................8 7 P e r c e n t a g e s o f d r o p o u t s b y p r e v e n t i o n p r o g r a m a n d SMOKING STATUS AT BASELINE.................................................................................. 8 8 P r e v a l e n c e r a t e s o f c i g a r e t t e u s e o v e r t i m e c d .......................HO P r e v a l e n c e r a t e s (%) o f l i f e t i m e a n d c u r r e n t ( p a s t 3 0 DAYS) SMOKING AMONG EIGHTH GRADE ADOLESCENTS WITH DIFFERENT ETHNICITY BETWEEN CALIFORNIA. NATIONAL. AND F l o r i d a s a m p l e s ................................................................................................................0 2 In t r a c l a s s c o r r e l a t i o n c o e f f i c i e n t s dCC's) a n d d e s i g n EFFECTS FOR SMOKING RELATED VARIABLES................................................0 0 M e a s u r e m e n t (p) p a r a m e t e r e s t i m a t e s o f l a t e n t s t a t u s i n RELATION TO MEASURED VARIABLES r .................................................................9 8 U n c o n d i t i o n a l p r o b a b i l i t i e s (o p a r a m e t e r ) o f l a t e n t STATUS AND PREDICTED NUMBER OF SUBJECTS FOR ENTIRE SAMPLE AND BY GROUP.................................................................................................... 9 9 T r a n s i t i o n p r o b a b i l i t i e s u p a r a m e t e r ) b e t w e e n l a t e n t STATUSES FOR ENTIRE SAMPLE AND BY GROUP ........................................ 1 0 1 O m n i b u s t e s t f o r u n c o n d i t i o n a l p r o b a b i l i t i e s (5) a n d TRANSITIONAL PROBABILITIES ( t ) BETWEEN GROUPS ......................... 1 0 4 T r a n s i t i o n p r o b a b i l i t i e s b e t w e e n l a t e n t s t a t u s e s b y GENDER......................................................................................................................................1 0 6 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. IX T a b l e 5 . 11 T a b l e 5 .1 2 T a b l e 5 .1 2 T a b l e 5. u T a b l e 5 .1 5 T a b l e 5 .1 6 T r a n s i t i o n - p r o b a b i l i t i e s a m o n g l a t e n t s t a t u s e s b y ETHNICITY...............................................................................................................................1 0 8 C o m p a r i s o n s b e t w e e n p r e d i c t e d a n d a c t u a l p r o p o r t i o n s OF SMOKING STAGES FOR ENTIRE SAMPLE ( N = 2 . 0 5 o ) .......................... 1 0 9 O d d s r a t i o s o f l a t e n t c l a s s m e m b e r s h i p i n r e l a t i o n t o PREDICTORS............................................................................................................................1 2 0 C o m p a r i s o n o f p a r a m e t e r e s t i m a t e s a n d m o d e l fit b e t w e e n THE S A S P R O C TRA-J PROCEDURE AND THE MPLUS PROGRAM 1 2 3 P r o g r a m e f f e c t s o n t h e t r a j e c t o r i e s o f l i f e t i m e s m o k i n g MEDIATED THROUGH CHANGING INTENTION TO SM OK E....................1 2 6 G E E MODELS FOR REGULAR AND ADDICTIVE SMOKING OYER TIME ...........................................................................................................................................................1 3 0 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. LIST OF FIGURES X FIGURE 3.1 FlCrl’RE -1.1 F i g u r e 1.2 F i g u r e -1.3 F i g u r e 3. i F i g u r e 3.2 F i g u r e 3.3 F i g u r e 3.-1 F i g u r e 3.3 F i g u r e 5 .6 F i g u r e 3.7 F i g u r e 3 .8 F i g u r e 3 .9 Co n c e p t u a l m o d e l o f p s y c h o -s o c i o -b e h a v i o r a l i n f l u e n c e s ON SMOKING PROGRESSION..........................................................................................42 H y p o t h e s i z e d m o d e l f o r s m o k i n g t r a n s i t i o n a c r o s s s t a g e s 6 0 Co n c e p t u a l l a t e n t g r o w t h m i x t u r e m o d e l o f s m o k i n g ....7 0 T h e o r e t i c a l m o d e l s o f m e d i a t i o n (A) a n d m o d e r a t i o n (B) 73 C h a n g e o f s m o k i n g s t a g e s ...................................................................................93 In c r e a s i n g t r e n d o f r e g u l a r /a d d i c t i v e s m o k i n g .........................9-1 S m o k i n g T r a j e c t o r i e s f o r t h e F n t i r e S a m p l e ..............................112 S m o k i n g T r a j e c t o r i e s b y P r e v e n t i o n ('.r o u p s ...............................113 S m o k i n g T r a j e c t o r i e s b y c. k n d e r ................................................................ 1 15 S m o k i n g T r a j e c t o r i e s by E t h n i c i t y ..........................................................117 M u l t i l e v e l m e d l a t i o n a l m o d e l f o r t e s t i n g t h e p r o g r a m e f f e c t o n s m o k i n g t r a j e c t o r i e s m e d i a t e d t h r o u g h CHANGING THE MEAN LEVEL AND GROWTH RATE OF INTENTION TO SMOKE...........................................................................................................................................125 P r e v a l e n c e r a t e s o f l i f e t i m e s m o k i n g b y t h e l e v e l s o f FRIENDS’ SMOKING (FS) AND REFUSAL SELF-EFFICACY (RSE)... 127 L a t e n t g r o w t h c u r v e m o d e l f o r t e s t i n g i n t e r a c t i o n s BETWEEN FRIENDS’ SMOKING AND REFUSAL SELF-EFFICACY ON LIFETIME SMOKING............................................................................................................128 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. ABSTRACT XI The Tobacco Program and Policy Trial (TOPP) was a two-group randomized trial that compared a tobacco policy support program with a policy as usual control condition in 19 schools in Southern California (n= 2.053 7th graders). The objectives of the present study were to examine patterns of adolescent smoking progression, to determine relative impor tance of the prevention program, intrapersonal, interpersonal, and demo graphic variables, and mediating and moderating mechanisms among these factors in predicting progressions of adolescent smoking across three waves of measurement. The longitudinal sample comprises 48.5% males, 40.7% Whites. 2.4% African Americans. 29.9% Hispanic. 1(5.7% Asians, and 10.3% others. The results showed that there was a significant increase in preva lence rates of lifetime smoking (16% vs. 30%) and current (past 30 days) smoking (3.2% vs. 7.2%) from 7th to 8th grade. The prevalence rates of smoking were higher in males. Hispanics, African Americans, and in the control group. Overall 89% of adolescents were in a non-smoker latent status at the beginning of 7th grade, 82% at the middle of 7th grade, and 76% at the middle of 8th grade, indicating a rapid progression of smoking. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. Three latent classes of adolescents were identified based on their smok ing trajectories: stable non-smokers (69.2%). slow escalators (28.4%). and stable smokers (2.4%). Decreased intention to smoke significantly medi ated 50% of the program effect on reduced progression of lifetime smok ing. Friends' smoking interacted with refusal self-efficacy on both mean levels and growth rates of lifetime smoking. Finally, significant linear and quadratic time effects on adolescent regular and addictive smoking were detected in the present research. Male gender, intention to smoke, low academic grades, parental smoking, and friends' smoking were the most important factors associated increased risk of adolescent smoking progression, and regular or addictive smoking, while prevention program and high socioeconomic status were associated with decreased risk of smoking progression. The findings suggest that intervention aimed at enhancing school non-smoking policy may reduce the probability of both initiating and pro gressing in smoking. Factors of intention to smoke, friends’ smoking, and refusal self-efficacy should continue to be emphasized in adolescent smok ing prevention programs. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 1 * CHAPTER 1 INTRODUCTION Tobacco use remains the leading preventable cause of death and disease in the United States.-1 1 Cigarette smoking has been linked to lung cancer, coronary heart disease, and stroke.1'7-01 Although these diseases typically do not occur until adulthood, the negative effects of to bacco use may appear as early as adolescence. For example, smoking by adolescents is associated with an increased risk of early atherosclerotic lesions, a precursor to coronary artery disease. Additionally, tobacco use by adolescents increases susceptibility to adult smoking and subsequent disease, especially because most adolescents who smoke regularly con tinue to smoke throughout adulthood.-0’ ’ Thus, effective prevention ef forts that target adolescent tobacco use are needed to avert the develop ment of adult diseases that begin with tobacco use in adolescence. Despite increased public knowledge about the adverse health ef fects of smoking, which may contribute to the significant decline of cur rent smoking among adults.207 adolescents still experiment with tobacco use. Nationwide, the prevalence rate of tobacco use among adolescents R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. increased during the 1990s.27 !S The incidence rate of first tobacco use in creased by 30% and the rate of first daily tobacco use increased by 50% during 1988 - 1996.2 S Tobacco use among White. Hispanic, and Asian adolescents is increasing more rapidly than that among African American adolescents.2 (,r’ In California, the prevalence of tobacco use among adoles cents was stable from 1990 to 1993 but then increased through the end of the decade.172 While there has been a slight decline nationally in adoles cent smoking in 2001. the major progression of new smokers has been and continued to be adolescents.lso Thus, identifying different patterns of adolescent smoking progression is an important issue because such het erogeneity has important implications for prevention and research. Smoking uptake in adolescence has been conceptualized as pro gressing through a sequence of developmental stages.7 1 ’1 2 1 Social, psycho logical. behavioral, and biological factors may influence this progression. Most current prevention programs either focused on one of three strate gies: universal or primary prevention.70 I,i;i 10 7 intervention with high risk youth who were smokers,201 and to a lesser extent, smoking cessation pro grams for smokers who were caught smoking or who wanted to quit.71171 1 7 1 However, none of these addressed different patterns of growth or progression of smoking. Prevention programs that could ad- R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 clress these progressions would constitute more components that are more tailored to a youth’s needs and readiness to prevent future smoking. Leventhal and Cleary1 -’* proposed that developmental history for the individual smoker moved in steps through the stage of preparation, initiation, becoming a smoker, and maintenance of smoking. Similar stages of smoking were suggested by Flay et al.7 ,; and Stern et al.1 !,!' The following summarizes the basic stages based on Flay et al.7 ,! and Surgeon General's Report.-1 1 '’ Stage I. Pre-contemplation. Adolescents have never smoked in this stage. They may have no desire to start smoking or have never thought about smoking. Stage 2. Contemplation or preparation or susceptibility. Adoles cents begin to think about smoking and intend to smoke in this stage. They are forming beliefs and developing attitudes about smoking ciga rettes. Exposure to media messages or role models may play an impor tant role in this stage, in which adolescents are susceptible to peer pres sure. Stage 3. Initiation or tried cigarettes. Adolescents try the First cigarette or first few cigarettes. This stage was considered the potential step to experimental stage, although many youngsters decided not to con R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 4 tinue with their experimentation.1 -* Peer pressure may be a stronger predictor than family influence in this stage. Stage 4. Experimentation or becoming a smoker. Adolescents gradually increase the frequency and number of cigarettes in various situations. In this stage, youngsters may develop a self-image as a smoker. They may receive minimal pleasure from smoking. They em phasize much on the positive aspects hut little negative aspects of smok ing.1 1 ''1 However, they are not totally committed to smoking in the future. Perceived positive benefits of smoking and reduction of harsh cues (e.g., burning, roughness, bite of the heat) may contribute to the continuation of smoking. Family members’ smoking may increase opportunities for adolescents to obtain cigarettes. Stage 5. Regular or established smoking. Adolescents progress to smoking on a regular, yet still infrequent basis. They smoke at least monthly or weekly, but they are not smoking heavily. Many regular smokers may not progress to heavy or addictive smokers. They may have smoked more than 100 cigarettes in their whole life. Tobacco use be comes increasingly regular in variety of situations such as every weekend or on the way to or from school. Stage 6. Addictive smoking or nicotine dependence. Adolescents smoke on a daily basis in this stage. They may experience addiction or R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. nicotine dependence and develop cravings, heavy daily use of cigarettes, and withdrawal symptoms. Adolescents, as much as adults, find it diffi cult to quit smoking at this stage.-01 In this stage, both psychological and biological factors contribute to the maintenance of smoking behavior. In empirical studies, the measures of smoking stages are primarily based on the definition of stages. For example, smoking more than one cigarette in one's whole life, but not having smoked in the last week has been used a measure for triers.™" Having smoked or tried a cigarette but not smoked 100 cigarettes in their whole life or smoked 10 or more cigarettes in the past 30 days has been used as a measure for experimen tal smoking.™ Having smoked more than 100 cigarettes in their entire life and smoked in the past 30 clays have been used as a measure for regular smoking.™ ™ -os Meanwhile, Flay et al.7 7 used reported smoking cigarettes in the past week to measure regular smoking. On the other hand, monthly use of cigarettes has been used to determine stages in other studies.i0 However, in a few studies empirical measures for estab lished smoking were used as same as the regular smoking.™ The measures of nicotine dependence or addiction have been devel oped for adult populations. These measures include the Fagerstrom Tol erance Questionnaire (FTQ),70 the Fagerstrom Test for Nicotine Depend ence (FTND),0 1 the Heavy Smoking Index (HIS),0- the Nicotine Depencl- R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 6 ence Syndrome Scale (NDSS).ls,i and the Stanford Dependence Index (SDI).17" ’ The major items in these instruments include the following as sessments: smoking frequency, time of smoking after waking, smoking when ill, difficulty refraining from smoking, smoking more heavily in the morning than other times of the day, cigarettes that one hates to give up, and nicotine levels of usual brand of cigarettes. The applications of nico tine dependence measures for adolescent populations may be plausi ble.1 - 1 1 1 but empirical studies using these measures to assess adolescent nicotine addiction are limited.1 v’1 [“•» Further research is needed to establish the validity of these measurements applied in adolescent popu lations. Although adolescent smoking uptake has been described as consisting of several stages, not all adolescents smoke or progress from trial smoking to addictive use. Various trajectories of smoking may be linked to different etiological pathways. An increasing number of studies with regard to smoking progression (transition, growth curve, or trajec tory) in adolescence have been found in the literature. New advanced statistical methods have been developed, which provide enormous means to assess adolescent smoking progression. Latent growth mixture modeling captures heterogeneity in growth by identifying discrete classes or mixtures.1 ni ,s This type of modeling can be used to empirically cluster participants’ trajectories of smoking.1 - 1 1 Latent transition R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. ticipants’ trajectories of smoking.1 2 1 1 Latent transition analysis™ has been used to test models of stage-sequential development of multiple drug use including cigarette use.IS s:! The development of latent growth curve modeling enables researchers to test the mediator and moderator effects among latent variable.*’212* ’ 1 lf) Particularly recent advances in random effects or multilevel modeling and generalized estimating equa tions (GEE) modeling for both continuous and binary longitudinal data provide a superior approach to analyzing both intra- and inter-individual changes over time and evaluating prevention program effective ness.2 1 !";12!,n'’1 1 ,2 2"’ However, the application of these advanced methods in the research of smoking progression and/or prevention is still in its in fancy. Factors related to cigarette smoking are varied, and may differ de pending upon the specific developmental stages of life span and stages of drug use.'s!U51 Several intrapersonal (psycho-behavioral) and interper sonal (social) factors have been identified in relation to onset of adoles cent tobacco use.5 1 Intentions and favorable attitudes toward cigarette use predict trial use.H 1 2 1 7 5 1 200 Refusal self-efficacy may predict smoking initiation and e s c a l a t i o n . 122 Peer and parent use appear to predict some progression beyond trial use. ”8-"-1 0 1 •1 0 (J-1 :Jy -1 '|;,-2 H 2 Family and school bonding are associated with onset.I1* ’II'S15517S Social influences mav also R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 8 have indirect effects on smoking initiation and/or escalation mediated through intrapersonal factors.7sl:!- Interactions among social and in trapersonal factors may promote smoking onset and escalation.197198 However, the predictors of higher-level smoking such as addictive smok ing and progression across different smoking stages are not altogether clear. Jessor’s Problem-Behavior Theory.I< ir’ the Social Learning Theory,9 the Theory of Planned Behavior.- and the Family Interactive Theory-1 provide best theoretical bases for the predictions of initiation and further smoking transitions. Meanwhile, smoking prevention programs that have countered these personal and social influences have shown suc cesses in preventing onset and progression from trial to more regular use.-1,i7,iH 79 I'i" ’ 1 ,1 7 However relatively little is known about the capacity of such programs to offset progression to addictive use. A study of the rela tionships of demogi-aphic, intrapersonal, and interpersonal factors to the progressions of adolescent smoking beyond trial use would help inform future prevention programs. In the present study, we First examined the transition patterns and probabilities of adolescent smoking across different stages. Second, we empirically identified adolescent smoking trajectories and related risk factors. Third, we tested escalation to addictive smoking two ways, In R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 9 rapid escalation to more amount/frequency of smoking, and validation by the items that ask if subjects smoked within 30 minutes of waking. In fluences of demographic, intrapersonal, and interpersonal factors on regular and addictive smoking were examined over time. Fourth, we tested moderator and mediating effects among intrapersonal, interper sonal, and demographic variables, and prevention program on adolescent smoking progression. Finally, using a large data set with complete in formation and great ethnic diversities, as in California, different pat terns, transition probabilities, and trajectories in diverse ethnic adoles cents were examined. This paper is organized into six chapters. In chapter two. we re viewed published longitudinal studies regarding smoking stages, devel opmental rates, transition probabilities, trajectories, and associated pre dictors. Four categories of studies were summarized: progression across stages, transition rates, latent growth rates, and multiple trajectories. Study design, analytic methods, and major findings were examined. Chapter three provided a descriptive overview of study design, sub jects, and major measured variables. In particular, the Tobacco Program and Policy Trial (TOPP),H i- from which the data for the present study were drawn, was reviewed. The present study utilized the data from the Project TOPP to do extended secondary data analyses. The original spe R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 10 cific aims of the Project TOPP were primarily to test the direct, indirect, and contextual effects of policy interventions, changed social norms, and social variables on reducing adolescent tobacco use. The ultimate objec tive of the present research was to further explore and utilize the data from the Project TOPP. In chapter four, new statistical methods for analyzing longitudinal data, especially those most suitable for assessing the changes of adoles cent smoking behaviors, were summarized. These advanced statistical methods include latent transition analysis (LTA). latent growth curve modeling (LGC). latent growth mixture modeling (LGM), and generalized estimating equations modeling (GEE). Major theoretical background and their applications were reviewed for each method. In chapter five, we demonstrated the results of adolescent smoking progression patterns and associated risk factors using new advanced sta tistical methodologies. Comparisons of transition probabilities and het erogeneity of trajectories between genders, ethnicities, and prevention groups were made. The relationships of intrapersonal, interpersonal, demographic variables, and the prevention program to different trajecto ries were examined. Mediating and moderating mechanisms of intraper sonal (e.g., intention to smoke and refusal self-efficacy) and interpersonal factors (e.g.. friends' smoking) on adolescent smoking progression were R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 11 also tested. Chronological trends and patterns, and risk factors for addic tive smoking behaviors were identified. In chapter six. we emphasized the implications of our findings about transition rates and heterogeneity of growth trajectories in under standing the patterns and pathways of adolescent smoking progression. The program effects of the Project TOPP on altering the pathways of ado lescent smoking progression and deterring uptake towards addictive smoking behaviors were evaluated. Finally, we pointed out the study limitations and strengths, and drew conclusions of our study. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 1 2 CHAPTER 2 ADOLESCENT SMOKING PROGRESSIONS: A REVIEW In this review, longitudinal studies that addressed progressions across discrete stages of smoking or in continuous fashion were included. Longitudinal or cohort studies may provide stronger evidence for the de velopmental nature of smoking progression. The key words used in lit erature search were "progression." "change,” "transition." "growth curve," "trajectory." "development," or "longitudinal pattern" in combination with "tobacco use." "cigarette use," "substance use." or "smoking." The search databases included "Medline" (medicine and health science). "Psyclnfo" (psychology), "Eric" (education), and "SocialSciAbs" (social science). The literature search yielded 31 studies meeting the criteria of longitudinal studies and key words. The articles in this review represent studies in the United States. New Zealand, Australia, the United Kingdom, Can ada. and Sweden. The present review extended a prior review on smok ing stages11 1 in at least three ways. First, we focused on studies with a longitudinal or cohort design. Second, we included studies with regard to R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 13 the estimation of transition probabilities across smoking stages. Third, we reviewed recently published studies regarding smoking growth rates and smoking trajectories, particularly those using latent growth mixture modeling to identify subgroups of individuals with different growth tra jectories. Based on the types of study design and analytical methodologies, four categories were classified (Table 2.1). Studies in Category 1 ad dressed patterns and predictors related to progressions of smoking be tween two stages. Studies in Category 2 addressed the probability of transitions across different stages of smoking. Studies in Category 3 ex amined average growth rates of tobacco use over time and related risk factors. Studies in Category -1 determined sub-classes of subjects with distinct developmental trajectories. Table 2.1 C ategories of studies on sm oking progression C ategory Type of C hange in Stages A nalytical M ethods C ategory 1 C ategory 2 C ategory 3 C ategory -I P rogress b e tw e e n sta g e s T ran sitio n probabilities G row th C urve M u ltip le T rajectory L ogistic regression a n a ly sis; Cox proportional h a za rd model. L a ten t tran sition al a n a ly sis; M ark ov chain a n a ly sis. L a ten t grow th cu rve m od elin g. F in ite m ixture gro w th cu rv e m odeling: L aten t class grow th cu rv e m odeling R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 14 Within each category, age or grade, gender, site of study, sample size, study design, analytical methods, statistical packages, patterns of smoking progression, and predictors or risk factors were evaluated and summarized (Table 2.2). 2.1 Analytic Methods In most of the studies in category 1, odds ratios or hazard ratios obtained from multivariate logistic regression analysis or Cox propor tional hazard models were used to measure the effects of risk factors on progression of smoking stages. Other multivariate analytic methods such as multivariate discriminate function analysis were also used in this category.8 Multivariate proportional odds models were used in one study.1 8 1 The proportional odds models provide cumulative odds ratios modeling the probability of being in any higher category on the smoking index given a baseline category. With an ordered dependent variable, these models have the advantages of retaining information that would be lost by combining the data into two arbitrary groups, as one does when using logistic regression. Random-effects logistic models were used to ac count for the design effects in studies with repeated measures.7 7 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 1 5 Table 2.2 S u m m ary of studies on sm oking stages/progressions A uthor. ACe/grade. gen Study design, an a Progression o f Predictors/ year der. site, and sam ple size lytic methods, and statistical package smoking stages Kisk factors C ategory 1 - P rogression between S ta g es S k a r a et Aged l a - 10 y e a rs . L o n g i tu d in a l study; P rogression fro m C i g a r e t t e p r e v a le n c e al , :" 1 2 w a v e s (fall 100a Logistic reg ressio n e x p e r i m e n t a l to r e g u e s t i m a t e . F am ily c o n 2001 s u m m e r 10001. S o u t h e r n C a l if o r nia. X = 1 . 0 U a n a ly s is lar cigarette use flict; I n te n tio n to use. P er c e iv e d stress. S e n s a ti o n s eek in g S a r g e n t G r a d e 1 t h r o u g h P ro sp ective cohort X e v e r s m o k e r -'nun- H i g h e r g r a d e in e t al..IM 1 1 1 a t b a s e lin e tn study: susceptible. X e v e r school. P e e r s m o kin g . 2 00 0 S e p t e m b e r 1000. 2 ,J s u r v e y in S e p t 1007. :L ! s u r v y m m a y 1008. Ver- m nnt. N’= 180 O d d s ratios {logistic modi'll. C u m u l a t i v e 1 od ds ra tios { propor tional o d d s modelsi s m o k e r /susceptible. Puffer: 2-10(1. not c u r r e n t smoker. 2-100/ c u r r e n t s m o k e r s R ecep tiv ity to c ig a r ette p r o m o tio n s Siegel a n d 12-1 a y e a r s at L o n g i tu d in a l study. Progression to e s t a b A n t i s m o k i n g media. B ie n tT . 1** initial in terv ie w . Logistic reg ressio n lished sm o k in g B as elin e s m o k ing . 2000 Follow-u p 1 v e a r s a n a ly s is (SASi (smoked m o re 100 A dult s m o k e r . F riend later. M a s s a c h u c ig arettes m li fetim ei s m o k e r setts. X = a0 2 W a n g f t 12-18 y e a r s at C o ho rt studv. I’wo progression.' School p erf o r m a n c e . al tu n e 1. la -22 y e a r F a c to r a n a h -is. M u lti v a r ia te logistic reg res s io n a n a ly s is a) F rom n o n s in o k '-r to ex p e ri m e n t e r . hi F rom n o n s m o k e r p erce iv ed teachers' 1 !)!)!> at tim e 11. T w o time points w ith th r e e - v e a r lag. the s m o k in g , frie n d s ’ a n d p a r e n t s s m ok in g, p erce iv ed f r ie n d ' t ’S n a t io n a l s a m to reg u lar s m o k e r ap p ro v al, beliefs, a n d ple. N= 10.12 b eing W h i te W e s t ft Aged la at baso- L o n g itu d in a l studv. I ’p ta k e of r e g u l a r F r i e n d ' sm o k in g al . 1000 line. F o u r o c c a sions age of 10. C h i- s q u a r e test. s m o k in g in 1 p e r io d s ‘ 1 ■ ”>-1 (i. l(i-18 18-21. p r e d ic te d u p t a k e of r e g u l a r s m o k in g in is. 21. a n d 2d M u lti v a r ia te logistic a n d 21 -2:1 years). m id -ad o le sce n ce a n d over 8 - y e a r period. reg res s io n analvsis. R egu lar s m o k i n g d o u e a rl y a d u l th o o d W est of S co tlan d . bled betw e en I o i 1 1"") N = 1,000 a n d 2.1 idli'L) y e a r s D ishio n f t G r a d e 1 ( m e a n C oh o rt s e q u e n tia l F o u r stages a b Low SKS. al 1000 age 10 vears). design. stainer. e x p e r i m e n t a l . Low- social preference. I n te r a c ti o n b etw e en S e v e n w a v e s of a s s e s s m e n t. Hoys. D iscrete tu n e event h isto ry a n a ly s is {logistic reg ressio n a n a lysis i p a tte r n e d , a n d daily. R a te s of initiation Pacific N o r t h w e s t . increase at ( Irade 8 p a r e n t a l s u b s t a n c e C ohort 1 lN = 102). Cohort 2 ( N = 1 0 ll a n d p eak at Grade!) Use a n d low ac a d e m ic a c h i e v e m e nt D is te fan Aged 12-18 y e a r s L o n g i tu d in a l s tu d y Pro gression from M ale a n d fem ale o t al ,'"s at first s u r v e y in w ith m u l tis ta g e n e v e r sm o k in g to frie nd s w h o sm oke. 100S 1080. a n d la-2 2 s a m p l e design: e x p e ri m e n t a ti o n , fro m S u sc ep tib ility to fu y e a rs at T u n e II. the U S n a t io n a l s am p le. X = 7 .000 Logistic reg ressio n a n a ly s is ( S l ’D AA X i e x p e r i m e n t a ti o n to e s ta b lis h e d s m o k i n g t u r e s m o k i n g . P a r e n t s ' a t t i t u d e to w a r d f u tu re s m o k i n g o f teen F lin t ft Kxpe r u n e n t e is L o n g itu d in a l s tud v Progression from Race ( W h ite vs black). al.. s" 1008 aged 12-18 y e a r s at first s u rv ey , la- w ith m u l tis ta g e s a m p l e design. L o e x p e r i m e n t a ti o n to r e g u l a r s m o k in g ■Sex (m a le vs female). 22 y e a rs a t T i m e gistic reg res s io n F r ie n d sm o k in g . II. th e U n ite d S tates . X =2. 107. a n a ly s is (SL'DA A X a n d SAS). P a r e n t ’s a t tit u d e R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 16 Table 2.2 (C ontin ued) A uthor, Age/grade, gen- S tu d y design, ana- Progression of P red ictors/ year d er, site, and lytic methods, an d sm oking stages Risk factors sam ple size statistical package B r e s la u et A g e 2 1 - to .10- L o n g i tu d in a l s tu d y . P r o g r e s s i o n to daily H is to r y o f m a j o r d e al 1!>!)* ye ar). I follow- u p s S o u t h e r n M ic h i g a n . X=P7 1 o d d s ratios (logistic model): H a z a r d s r atio s (Cox niodels) s m o k i n g p r e s s i o n a t b a s e l in e l-'lav et G r a d e 7. T l and L o n g i tu d in a l r e T h r e e tr a n s i tio n s F r ie n d s ' s m o k i n g . al . " T 2 w ith a - y e a r search . X e v e r u s e r s to triers. c i g a r e t t e offers, i n t e n m s lag: Califo rnia. M u lti n o m ia l logistic re g r e s s io n analvsis. n e v e r u s e r s to e x p e r i m e n t e r s . a n d n e v e r u s e r s to re g u la r tions. sch o o l g r a d e , alcohol a n d m a r i j u a n a use. P a r e n t a l s m o k i n g N = 2.!M2 s m o k e r s a n d fam ily conflicts P a tt o n et A g e d 1 1 y e a r s at Ci ihort s t u d x . T w o tr a n s i tio n s P a r e n t a l d a i ly s m o k al , !" ItMIS b as elin e. S i \ w a v e s w ith <>- m o n t h interval. Victoria. A u s t r a lia. N'= l.P 17 Cox p r o p ortio nal h a z a r d s m odel (S A S a n d S ta t a ) X e v e r s m o k in g - a n y s m o k i n g - daily s m o k ing. Incident sm o k in g - c e ss a tio n - relapse ing. P a r e n t a l divorce. S m o k i n g at o u ts e t . M e t r o p o l i t a n school Pierce et A g e d 12-17 years L o n g i tu d in a l study. S m o k i n g u p t a k e from Rec ep tiv ity o f c i g a al m s at b aseline. -I y e a r s a p a r t . C ali fornia . N = I.7u2 • la ck k n ife p ro ced u re . A t tr i b u ta b l e risk. Logistic reg res s io n nil 1 1-susc eptible n ev e r s m o k e r s to e x p e r i m e n t e r s r e t te p r o m o t i o n s . N o n sig nifican t fam ily a n d p e e r in f l u e n c e s W ant: et Ageil 11 IS y ears L o n g i tu d in a l studv. T w o p ro g ressio n s M o t h e r s m o k e s . F a al . m r (baselin e). la-22 y e a r s iT u n e I I ). Gilds ratio - (SAS) n o n - s m o k e r s - e x p e r i m e n t a l sm ok ers. t h e r s m o k e s . O l d e r b r o t h e r s m o k e s . T h e L’.S n atio n al e x p e r i m e n t a l s m o k e r s F r ie n d s m o k e r (both s a m p l e . N=<i.ut!) - r e g u l a r s m o k e r s m a l e a n d f e m a le ) Choi et A ged 1 2 -IS y ears L o n g i tu d in a l s tu d v P r o g r e s sio n from F r ie n d s s m o k e . F’ain- al.. ( T A P S 1). F o u r w ith m u l tis ta g e e x p e r i m e n t a l sm o k in g dv m e m b e r s s m o k e . P>!)7 y e a r s b e t w e e n s a m p l e design. at T 1 to es ta b lis h ed C u r r e n t e x p e r i m e n t T A P S I a n d 11. the C S n a t io n a l s a m ple. \ = 2.liS 1 (puffed) M u ltip le logistic re g r e s s io n a n a lv s is ( S l 'l ) A A N i s m o k i n g at T2 dl ’. e r s it b a s e lin e , School p e r f o r m a n c e . No c o m m i t m e n t not to s m o k e Ary a n d 7>h. K) 0: g[-;nfe r s L o n g i tu d in a l study. C h a n g e s from n o n P e e r s m o k i n g p r e Biglan.' PIPS I initial test) ( T 1 > . 1 -v e a r follow-up (T2). I.ane C ou n tv. O r e g o n . N=S()I M A N O V A . H i e r a r chical m ultiple r e g r es s io n an a lv s is S tep w ise d is c r i m i n a n t function a n a l y sis (SAS) s m o k i n g tn sm o k in g d ic te d c o n t i n u a t i o n of s m o k i n g P e d e r s o n A ge 10 a t t u n e I. S L o n g itu d in a l. C h a n g e s b e t w e e n F e m a l e . O l d e r age. a n d I.ef- c o e .1 y e a r s a p a r t at t u n e 2; O n tario . D i s c r i m i n a n t f u n c tion analysis; C h a n g e score (-1. 1). - l i n e v e r s m o k e r s , c u r r e n t s m o k e r s , a n d P a r e n t 's S F S , P e e r s m o k i n g : P a r e n t a l l!)S(i C a n a d a . N = 2 . 2 l a tr i e r s s m o k in g : S t u d e n t p red ictio n . I n t e r v e n tion Cha.ssin (>-11"‘ g r a d e r s at L o n g i tu d in a l a n a l y T w o tr a n s i tio n s A j / e n a n d F i s h b e i n s et al . IPS t T l . O n e y e a r a p a r t at T 2 (7-12:l‘ g r a d e r s ) . M i d w e s t e r n cou n ty , N’=2.S IS. sis: D is c r im in a n t f u n ctio n analy sis: M u ltip le reg ressio n p r o c e d u r e s a l F r o m n e v e r s m o k ing to try in g smoke, hi F r o m tr y in g s m o k e to r e g u l a r s m o k i n g p r o x i m a l v a r i a b le s , -lessor a n d -lessor s dis ta l v a r ia b le s . S m o k i n g e n v i r o n m e n t R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 17 T ab le 2.2 (C on tin u ed ) Author. y ear Age/grade, gen der, site, and sam ple size S tudy design, ana lytic m ethods, and statistical package P rogression of sm o king stages Predictors/ Risk factors C ategory 2 - T ransition R a te M c C a r t h y e t al 20(11 M e a n a g e o f 12 8 ( S D = 8 7) y ears. Thre e a n n u a l in terv ie w . Los A n g ele s area. N = 2 5 1. C o h o r t stu d y . M a r k o v c h a in a n a l y sis. T r a n s i t i o n s b e t w e e n T l . T2. a n d T;1 a m o n g t h r e e c l as sificatio n s N o n - s m o k e r , i n t e r m i t t e n t s m o k e r , a n d "everyday'" s m o k e r N o t e x a m i n e d J a n s o n . 1" 1 1999 A ged 12 y e a r s at first test; a n d t h e n a t a g e s o f la. 10. 17. is. a n d 21. 2a. a n d (0 y ea rs. S w e d en . N= 122 hoys a ml 00 girls L o n g i tu d in a l study; Kxact te s t o f c o n t i n g en c y tables. C l u s t e r a n a lv s is i S L K I P N K R ) 1 s u b - g r o u p s : s ta y i n g a n o n s m o k e r , s m o k ing in t e n s e l y a n d c o n t i n u i n g to a d u l t hood. s m o k i n g less in te n s e ly , o r q u it tin g Age at initiation F e r g u s s o n & N o r w o o d . " ■ 199.5 A ged 10 v e a r s at baseline. F o u r w a v e s w ith tw o y ea rly interval C h r i s t c h u rch. New Z e alan d . N = 9 5 7 L o n g i tu d in a l stu d v. T r a n s i t i o n a m o n g . t h r e e s t a g e s nun- L a t e n t M a i k o v tran- , s m o k i n g , o ccasio n al sition a n a ly s is s m o k in g , a n d r e g u l a r s m o k i n g C ategory ■ '{ - drou th C u n e Not e x a m i n e d 1 lo f f m a n n C h il d r e n ag e d I 1 - A ccelerated lo n g i t u S ig n ifican t g r o w th Life e v e n ts . et al 1 1 y e a r s d u r i n g d inal design, tl row til rate in c i g a r e t t e u-e I’e e r use. F am ily a t 2 0 00 th e first year. I- y e a r s tu d y period. M i d w e s t e r n l.'SA. N = 0 a l c u r v e a n a ly s is 1 two- level h ie r a r c h ic a l lin ear model). (Ms = I I. p < 1)1) t a c h m e n t . M ale C h o u et 7'h g r a d e r s at L o n g i tu d in a l studv. S ig n if ic a n t g r o w th School type. al . 100S baseline. 0 m o n t h follow-up. 1 y e a r a p a r t for o t h e r 1 follow -up>. I n d i anapolis. N = 57 schools. M ultilevel m odel tlll.M 1) a n d la tent grow ill c u r v e a n a l y sis (KQSi rate ill p r e v a l e n c e r a t e i if c i g a r e t t e use ( M s - 1 1)5. p < 1)1) I n t e r v e n t i o n p r o g r a m A n d r e w s A ged 1 1-la y e a r s L i n g i t u d i n a l studv. G r o w t h r a t e in c i g a G e n d e r a n d age in s ) a n d D u n ( m e a n 12.581 a t L a te n t g r o w t h c u rv e rette u se = I I (p < A t t i t u d e t o w a r d c ig a can." lOOS first a s s e s s m e n t , a g e d la-1 9 v e a r s at fifth follow-up. Five a n n u a l a s s e s s m e n t s . O r e gon. N = 2 5 0 m o d e lin g (KQS). 001). r e t te s (positive, p < 0.5) D u n c a n et Age 1 1-15 y e a rs L o n g i tu d in a l study. S ig n if ic a n t g r o w t h m C h a n g e in alcohol use al .v ’ [90S at b aselin e. Five a n n u a l a s s e s s m e n ts . O reg o n. N=70:! (.17 1 m ales. .189 fem ales) L a te n t G r o w t h M o d eling (LQ S ) c ig a r e tt e (M s = I !<i. t = -l 590). w a s significantly r e lated to t h e d ev e lo p m e n t of c ig arette use R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 18 T ab le 2.2 (C on tin u ed ) A uthor, vear Age/grade, gen der, site, and __sam £le_size__ S tudy design, a n a lytic m ethods, and ^ j s t a ti s t i c a l jj a c k a g ^ Progression of sm oking stages P redictors/ Risk factors C h a s s i n et al . '■ 1 I !)!)(> H u et al ''' C u ld e r et al . " 2001 C h a s s i n et a I . :j 2000 Wills et al . 1000 C h a s s i n et al . 1001 I.abnuvic et a ! . .1-" 1001 1 1'WI2"' g r a d e r s ( m e a n age IT yrsi a t baseline. 2"'1 follow-up iv n u n g ad u lts , m e a n age 2d yrsi; dr 1 follow- up (adults, m e a n a g e 20 yrs). M i d w e s te r n . N = -I.Odd 7I!’ g r a d e r s at T I. T 2 ( I m o n t h - from T l ). Td. T I ( I y e a r ap a rt). S o u t h e r n California . N = d . 7 10 G r a d e 0 7 ( 1 1 -1 d y e a r s of age) at b aselin e. Ill 17 y e a r s ( 10lh/1 1,h g r a d e ) at sixth follow-up. K a n s a s City. N = d2d t i r a d e (>-12 at baseline. Age 2 I - 11 y e a rs at last w ave. Six w aves. M i d w e s t e r n slate. \ - 0 . 2 2 d 7 " ’ g r a d e i m e a n ag e 12 I years) at baseline, yearly in t e r v a l s oxer i- y e a r period Low er W e s tc h e s te r C o u n ty . New York. N=1.1S I 0 " 1 - 12'1 ' g r a d e r s ( m e a n ag e 1 I to 1 I 0 years). T 2 <7- S y e a r s ap a rt, "ad u lts '), Mnl- w e s t e r n enmity: N = I. lob A g ed 12 y e a rs at initial test;2 r e te s t s a t ag e s la a n d IS years. N e w J ersey: N = 100 C o h o r t - s e q u e n t i a l stu dy: la m g i t u d i n a l ran- d om -effects r e g r e s sion a n a ly s is ( M I X O I M 7 ! L o n g i tu d in a l s tu dv . R an d o m - e f f e c ts o r din al r egression m od e l (M IN O R ) P r e s e n c e o r absence o f c u r r e n t r e g u la r s m o k e r s (at least w e e k l y ) o v e r d time p o in t s S ignificant in c r e a s e in p rev alen ce o v e r t i m e i(l = Id. p < 0001) K x a m i n e d tim e effects on a d o l e s c e n t c u r r e n t s m o k i n g Category I - Multiple Trajectory L o n g i tu d in a l s tu d y L a t e n t g r o w th m i x t u r e m o d e lin g (M plusi, Logistic r e g r e s s io n i.Mplus) C o h o r t- s e q u e n tia l Stlldv. M i x t u r e m o d e lin g iSAS Proc Tl'A-It. D is c r i m i n a n t f u n c tion a n a ly s is C o h ort studv. C l u s t e r i n g an a ly s is . P a ir w is e c o m p a r i son. M u l ti v a r ia te disi rimiii.ant f u n c tion a n a lv s is iSA.S i L o n g i tu d in a l s tu d v. Logistic r e g r e s s io n an a lvsis. 2 • 2 m u l t i v a r i a t e a n a ly s is of v a r i a n c e (M A N O Y A . SAS) L o n g i tu d in a l s tu d y. M u l t i v a r i a t e r e g r e s sion an a lysis. C l u s te r analysis: D is c r im in a n t function a n a ly s is (SAS) Five d is tin c t p a t te r n s K arly rapid e s c a l a tors. la te m o d e r a te es c a la to r s , late slow es c a la to r s , stable light s m o k e r s , a n d stab le puffers Six tr a je c t o r y groups a b s t a i n e r , erratic s m o k e r , e a rly -table s m o k e r , late -ta ble s m o k e r , qu itter, and e x p e r i m e n t e r Five s u b - g r o u p s S ta b l e n u n u s e r , m i n i m a l e x p e r i m e n t e r . late starter, a n d tw o es c a la to r g r o u p s F o u r g r o u p s stable n o n - s m o k e r s ((> I:V’»), s ta b le s m o k e r s i, IS 2 ”u). a d u lt-u nset s m o k e r s IS 7#»l. an d n o n - p e r s i s t e n t adoles ce n t s m o k e r s (S (i'„) T h r e e n o r m a t i v e lon g it u d i n a l p a t t e r n s c h r o n ic a lly low levels o f use. c h r o m c a ll i h ig h levels o f use. s h a r p in c r e a s e in use b e t w e e n th e a g e s of la a n d IS P a r e n t a l s m o k in g . A d o le s c e n t s m o k in g . I n t e r a c t i o n b e t w e e n p a r e n t a l a n d a d o l e s c e n t s m o k in g . K d u catio n F r ie n d s ' s m o k in g . P a r e n t a l s m o k in g . G e n d e r . K th n ic itv N ot e x a m i n e d F r i e n d - a n d p a r e n t s s m o k in g , to l e r a n c e for d e v i a n c e , locu- of con tro l, p a r e n t a l s u p p o rt. h e a l t h beliefs, a d u l t e d u c a tio n . |> a r - e n t h o o d F r i e n d s s m o k i n g b e e r. N e g a t iv e life e v e n t s . A n g e r a n d h a n g o u t coping. \ alu e d i s c r e p a n c y P a r e n t s m o k in g , liquor. A c a d e m i c c o m p e te n c e . Kmution.nl s u p p o r t A d o le s c e n t s beliefs ill n e g a t i v e social c o n s e q u e n c e s of s m o k i n g a n d a c a d e m i c success. A d i d t s beliefs in the n e g a t i v e h e a l t h c o n s e q u e n c e s of s m o k i n g H i g h level o f p r e v io u s use. G r e a t e r e a s e of m a k i n g a n d fin ding frie n d s. C h r o n ic a lly h i g h e r s u b s t a n c e use a m o n g siblings. H i g h e r level of beh a v - m r a I u n d e r - c o n t r o l R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 19 In the studies of category 2. Markov chain analysis was initially used to characterize the dynamism/stability of population prevalence es timates of smoking status over three annual observations.11 ,1 The latent Markov models assume that the processes that generate changes in the subject’s smoking status over time conform to a first-order Markov proc ess in which the subject's history of changes up to time / + / is summ a rized by their status at time t. This implies that the only information re quired to predict status at time / + 1 is the subject’s status at the previ ous point of observation.7" ’1 1 ,1 In category 3. hierarchical linear models and latent growth curve modeling were conducted to examine the average growth rates, initial levels, and their correlates for cigarette use over time."’1 '7 Longitudinal random-effects regression analyses for dichotomous outcomes were per formed in two studies.!,!m The MIXOR analyses model the group aver age of all individual age-related changes in smoking prevalence from ado lescence to adulthood.< M For continuous outcome variables, the multilevel modeling and latent growth curve modeling have been found to yield similar results in modeling individual growth.1 0 Studies in category 4 used latent growth mixture model that cap tures heterogeneity in growth by identifying discrete classes or mix tures1 IS empirically.5 2 1 1 Cluster analysis that was based on the similar- R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2 0 itv in absolute level was performed to group individuals with similar lon gitudinal tobacco smoking patterns over time.1 -0 Multivariate analysis of variance (MANOVA) was performed to compare the differences of smoking among 4 sub-groups: stable non-smokers, stable smokers, adult- onset smokers, and non-persistent adolescent smokers.r> In comparison with the studies in category 3. which modeled individual differences in the means and variances of latent growth factors in a continuous and av erage fashion, the studies in category 4 identified heterogeneity in growth patterns in a discrete, rather than a continuous fashion. 2.2 Sm oking Patterns Similar patterns of smoking progression from early adolescence to young adulthood or adulthood were identified in the most recent studies. Colder et al.1 1 identified five distinct longitudinal patterns of smoking progression. A group of early rapid escalators exhibited early escalation and their smoking increased rapidly after age 13. Another class, late moderate escalators, smoked very lightly until age 14 and they exhibited moderate escalation. Other three classes (late slow escalator, stable light smokers, and stable puffers) showed light smoking through out the study period. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2 1 Chassin et al.1 2 identified six groups with the distinct trajectories of smoking progression. The early stable group was characterized by a relatively early age of smoking onset, with an onset in the middle school year (age 12-13 years). They achieved a high level at a relatively early age (age 18-19 years) and they remained stable through the course of study. The late stable group had a later onset of smoking and had lower maximum level of smoking. They achieved this maximum level at age 24 and stayed stable at that level. The experimenter group had a relatively early age of onset but never escalated to heavy smoking. The quitter group reached their maximum smoking score between 20 and 21 years of age. and then began to decline, and no smoking was reported after 25 years of age. The erratic group was characterized by fluctuations around a group mean that shows an overall increase over age. and in general had its highest smoking rates at 23-25 years of age. The stable abstainer group who reported no smoking at any wave of measurement was identi fied as well. By using cluster analysis, different patterns of growth trajectories were identified. The Sweden study10* showed that light smoking (up to 6 cigarettes/day) did not remain stable after adolescence. Typical develop ment patterns from age 15 to age 36 included staying a nonsmoker, smoking intensely (>10 cigarettes/day) and continuing into adulthood. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. smoking loss intensely for some periods, and smoking intensely and quit ting before age 36. Differences in age at smoking initiation were related to later habitual smoking only when participants reported initiation had occurred after age 12. Meanwhile. Wills et al.-1 '1 detected five clusters. Stable nonuscrs (50%) showed almost complete nonuse of cigarettes at all three measure ment points. The minimal experimenters (26%) had scores representing a minimal level of cigarette use at all three measurement points. Late starters (14%) showed experimental levels in 7th and 8th grades but in creased in 9th grade. The escalator 1 (6%) and escalator 2 (4%) showed elevated cigarette use at 7th grade and increased in use throughout the study period. Similarly, three normative longitudinal patterns were identified in the study by Labouvie et al..1 -" namely, subjects exhibiting chronically low levels of cigarette use. those exhibiting chronically high level of use. and those exhibiting a sharp increase in use between the ages of 15 and 18 years. To examine the antecedents of smoking progression, four groups were constructed by an algorism based on their smoking status during adolescence and adulthood -- stable non-smokers, stable smokers, adult-onset smokers, and non-persistent adolescent smokers.1 '1 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2 3 2.3 Risk Factors Risk factors associated with smoking progression were identified from this review and were summarized in three major categories: in trapersonal or psycho-behavioral, interpersonal or social, and demo graphic factors (Table 2.2). 2.3.1 Intrapersonal or psycho-behavioral factors Intrapersonal or psycho-behavioral variables that were identified to predict smoking onset or escalation included attitude toward smoking, intention to smoke, health beliefs, depression or perceived stress, and school performance. 2.3.1.1 Attitude tow ard sm oking Positive attitude toward smoking has been shown to be a risk fac tor for young adults to become experimenters from non-smokers. Changes in attitude increased likelihood of future smoking such that more positive attitudes was associated with increased involvement with cigarettes.1 ™ The adolescent’s attitude significantly predicted both the mean level and growth rate of smoking such that adolescents with a more R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 24 positive attitude toward cigarettes at the first assessment were more likely to have higher levels of use at the second assessment. These ado lescents were more likely to increase their cigarette use from early time to later life at a faster rate than the adolescents with a less positive atti tude toward cigarettes.’ ’ More deviant attitudes and lower self-control were related to escalation of adolescent smoking.1 -'1-1 '’ 2.3.1.2 Intention to smoke Smoking intentions predicted onset of smoking,8 transition form trying to experimental use,1 7 7 7 and transition from experimentation to regular use.1 1 1 1 Susceptibility to smoking has been a strong predictor of progression from never-smokers to experimenters.")8 2.3.1.3 Health Beliefs Knowledge of harm and beliefs in peers’ caring about staying off cigarettes were independently associated with progression to regular smoking.80 Positive beliefs about smoking were significant predictors of young adults' smoking progression from nonsmoking to regular smok ers.-00 Smoking progression was preceded by normative beliefs and health beliefs that were more positive about smoking.1 ’ ’ Sensation seek ing predicted transition from experimentation to regular use.1 0 1 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2.3.1.4 Depression and perceived stress Depression scores were significant predictors of smoking progres sion from nonsmoking to regular smokers.-09 The significant association between progression to daily smoking, history of major depression, and current major depression was also identified.-- Perceived stress predicted transition to regular use.1 9 1 2.3.1.5 School perform ance Low academic achievement significantly predicted transition from trial to experimental use.77 Self-rated low school performance was a risk factor for young adult to become experimenters1 9 '9 and it was significantly associated with progression to established smoking.5 9 Not liking school and missing school were significant predictors of smoking progression from nonsmoking to regular smokers.-"9 2.3.2 In terpersonal or social factors Interpersonal or social variables were significantly related to ado lescent smoking progression in that more smoking parents, more smoking friends, and higher perceived prevalence of smoking in the community R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2 6 were preceded to elevated smoking use.'"’ Mass media and stressful life events were also found to predict smoking progression. 2.3.2.1 Friends* sm oking Friends' smoking seems to predict progressions of smoking across different stages. For example, the number of friends who smoked was a significant predictor of smoking progression from non-smokers to regular smokers.s -"" Having at least one female or male best friend who smoked was a significant predictor of progression from experimentation to estab lished smoking.■’ ,s-sn Friends' smoking also predicted transition from trial to experimental use, but not from experimental to regular use. Friends’ approval of smoking and cigarette offers by friends significantly predicted transition from trial to experimental use.7 7 Adolescents who perceived friends' approval were more likely to initiate and experiment with smok ing.-"'' There were gender differences in friends’ influences on adolescent smoking progression. For male adolescents, having one or more male or female best friends who smoked was significantly associated with smok ing progression from nonsmoking to experimental smoking. Having any best male friend wrho smoker was the only social factor that was signifi cantly placing a nonsmoking male adolescent at risk for becoming a regu- R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. lar smoker.1 7 -os In contrast, for nonsmoking female adolescents, only be ing with any best male rather than female friends who smoked was a significant risk for progressing to experimental smoking.-0* Several studies found no associations between friends’ smoking and adolescent smoking progression. For example, exposure to friend smokers did not appear to significantly influence which adolescents be gan smoking uptake process.17 0 There might be several possible explana tions for the different findings on the association between friends' smok ing and adolescent smoking progression. First, the measure of smoking within the past month (30 days) may underestimate the proportion of people who are in the early stages of the smoking uptake process. Sec ond, the influence of friends' smoking m facilitating and possibly encour aging adolescents to smoke may be the most apparent after first experi mentation. rather than influencing adolescents to experiment for the first time.170'IS 1 Third, it is possible that the peer effects have only a more proximal and short-lived influence on adolescent smoking behavior. 2.3.2.2 Family m em ber’s sm okin g West et al.-" found considerable variations in the patterns of asso ciation with uptake at different ages and revealed the 15-16-age period as being quite distinct. During this period, in addition to friends’ smoking, R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2 8 parents' and siblings’ smoking were also significantly associated with the uptake of regular smoking. However, the effects of family members' smoking, and friends' smoking as well, on the uptake of regular smoking appeared to ‘wear off over time, the former being rendered invisible, the latter being reduced in strength.1 7 '"’''-i'1 Parental daily smoking was sig nificantly associated with adolescent higher incidence of daily smoking, higher persistence of daily smoking, and much higher risk of relapse.1 7 ,s In contrast, some studies found that parents' smoking status was not a significant predictor of progression to experimentation among never smokers."’7', H although the interaction between parental substance use and low academic achievement reached significance."’7 Parental smoking significantly predicted the transition from experimental smoking to regu lar smoking, but not from trial to experimental smoking.77 Change in pa rental smoking was not found to be related to increased use of ciga rettes.1 7 ’9 Many studies have showed that friends’ smoking interacted with family members' smoking in relation to adolescent smoking progression. For example, levels of exposure to parents’ and friends’ smoking were strongly associated with progression to established smoking.1 9 l lrts The rate of progression to established smoking was highest among adoles cents who were exposed to smoking by both their family members and R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2 9 their best friends.1 9 Compared to early-onset smokers, late-onset smokers had fewer smoking friends and parents, and they perceived less smoking in their larger environments. Compared to stable never smokers, late- onset smokers differed on all antecedents except beliefs about the nega tive social consequences of smoking. Non-persistent smokers had fewer smoking parents than did those who remained smokers. Non-persistent smokers had more negative beliefs about the personally relevant health consequences of smoking than did their peers who smoked persistently into adulthood.1 * ’ 2.3.2.3 Prevalence estim ates The overestimation of smoking prevalence predicted transition from trying to experiment smoking, but not from experimental to regular smoking.77 In contrast, one study found that relatively low smoking prevalence estimates predicted the transition to regular use.1 1 ,1 Perceived number of teacher smoking and perceived friends’ approval-09 were sig nificant predictors of smoking progression from non-smokers to experi menters. R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 0 2.3.2.4 Stressful life events Utilizing a two-level growth curve model, Hoffmann et al.97 found that cumulative stress was significantly associated with the "growth" of adolescent drug use. As stressful life events increased in adolescence, a significant intra-individual escalation of drug use was observed. More over. increasing involvement in drug use was buffered by family attach ment. For adolescents with low family attachment and greater stressful life events, escalation of drug use tended to be rapid. Strong family at tachment also attenuated the effects of peer drug use on the escalation of drug use in adolescents. Greater life stress, lower parental support in daily life, more parental substance use, greater affiliation with peers who used substances were significant predictors of escalation of smoking.-1 '’ 2.3.2.5 Mass media Receptivity to tobacco promotions impacted tobacco uptake among susceptible never smokers and experimental smokers.1 7 - 1 S I Tobacco in dustry advertising and promotional activities influenced non-susceptible never smokers to start the process of becoming addicted to cigarettes.1 T - In contrast, exposure to television antismoking advertisements was found to reduce the likelihood to progress to established smoking.1 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 31 2.3.3 D em ographic Factors Demographic variables such as gender, ethnicity, and socioeco nomic status (SES) were found to predict adolescent smoking progression. 2.3.3.1 G en der Being a male gender predicted the transition to regular use.80101 Males had a slightly higher established smoking rate than females, though this was not significant.80 There were no significant differences in reported use between males and females in progression from never smoking to experimentation. In other studies, however, it was found that being a female gender and being older at baseline were significantly as sociated with increased use of cigarettes among adolescents. 1 5 0 Gender and age at initial study were not significant predictors of either the mean level or growth rate of smoking, indicating that these variables did not affect either the frequency of cigarette use at the later assessment or the growth in cigarette use in later follow-up assess ments.5’ ’0 R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 2 2.3.3.2 Ethnicity Large differences between African Americans and Whites in pro gression from experimentation to regular cigarette smoking were ob served in several studies. For example, being White was a significant risk factor to initiate smoking, to become experimenters, and to progress from nonsmoking to regular smokers compared to African American or Hispanic adolescents, to .-.s.so.^oo Minority youths were much more likely to progress only to susceptible smoking than non-Hispanic Whites, which suggested that the duration of the uptake process among minority groups was more extended, or that they began smoking progression at a late age.1 7 1 Ethnic groups differed markedly in their rates of progression to established smoking.1 1 1 2.3.3.3 Social economic sta tu s (SES) Low socioeconomic status and low peer social preference (e.g.. "number of kids whom I like or dislike to be friends with”) were signifi cant predictors for smoking onset and progression to regular smoking." Interestingly, it was found that adolescents from households where adults had more education were more likely to progress to established smoking than those from households where the responsible adults had less than a high school education.!!) R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 2.4 Theoretical and P ractical Issues 3 3 The smoking acquisition process has been conceptualized as con taining five distinct stages, i.e.. preparation, initiation, experimentation, habituation, and nicotine dependence.7 ,i The majority of current longitu dinal studies, especially those with short period of observation, focus on progressions from a lower level to a higher level of smoking stages. How ever. the practical measures for the theoretical stages varied substan tially among the different studies reviewed. Many risk factors were not found to be uniquely associated with a specific transition between stages and the results varied across studies. The lack of construct validity re lated to the measures of theoretical stages might reduce the likelihood of identifying specific predictors for smoking transition or progression and might cause the incomparability of the empirical findings. Theoretical development of smoking progression relies primarily on valid and reliable measures. To our knowledge, no psychometrically acceptable measure of smoking progression has been validated except the Stages of Acquisition Questionnaire on adolescents.m Although some measures for nicotine dependence (e.g., FTQ and FTND) have been validated among adult population7111 7 and these measures have been evaluated for the purpose of application to adoles R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 34 cent population.111 lr,ITr’ research on the reliability of existing measures and optimal nicotine measures especially for adolescents is still highly warranted. Furthermore, empirical studies on the smoking progression to nicotine addiction or dependence are rare in literature. However, it is still debatable whether adolescent smoking pro gresses through a series of distinct stages in light of the lack of consistent stage definitions. Willet-1! suggested that individual change should not be viewed as an incremental process, or as the difference between “before and after". Rather the individual change should be considered as a con tinuous process over time. An individual growth curve approach, model ing change over time, was recommended for future research.-M 2.5 E valuation of Prevention Effects As reviewed above, intrapersonal or psycho-behavioral factors, in terpersonal or social factors, and demographic factors were significantly associated smoking progression. Most studies only examined the main effects of these factors. Several studies examined interacting (or moder ating) effects among several intrapersonal, interpersonal, and demo graphic variables on the movement of smoking stages, ho wev er, these tests were primarily cross-sectional. No previous studies reviewed R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. have examined the mediating effects among intrapersonal and interper sonal variables, or prevention programs in particular. Further research that tests moderating and mediating effects among intrapersonal, inter personal factors, and prevention programs on smoking progression is needed. Factors such as school policy on smoking, policy enforcement on smoking, and family rulc-s on smoking have not been examined. Thus far. no studies about the effects of the prevention programs on smoking reduction have been reported by examining the longitudinal trends of adolescent smoking over time with one exception. Chou. Bentler, and P en t/1 0 examined the significant effect of a substance pre vention program among adolescents10" ’1 07 on the lowering of the average growth rate of adolescent smoking over four years. Further studies ex amining the short-term and long-term effects of smoking prevention pro grams on growth rates or transition rates or developmental patterns are highly demanded. During the development of adolescent smoking, some risk factors (especially psycho-behavioral and social factors) may be also changing. For example, intentions to smoke and perceived stress may increase dramatically if adolescents transit to a new school with new friends who may smoke. However, in all studies, risk factors were examined as time- invariant variables in multivariate models with one exception. Hoffmann R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 6 et al.o: examined the time-varying effects of stressful life events on pro gression of drug use. It is important to model some psychosocial risk fac tors as time-variant variables in relation to smoking variables because these predictors may show growth trends and trajectories as well. 2.6 Methodological Challenges The majority of the studies reviewed in this paper used odds ratios and hazard ratios to examine the effects of risk factors on dichotomized or ordinal-scale outcome variables. One of the advantages of these methods is that it can he used to identify the univariate or multivariate associa tions between a specific risk factor on a specific transition of smoking stages. Latent Markov Modeling and Latent Transition Analysis'1 0 are powerful methods for assessing the probabilities of transition among smoking stages. However, the assessment of transition probabilities is currently limited to that between two time points and no statistical pack ages are currently available for modeling associations among transition probabilities and related variables that may influence these probabilities across stages. Latent growth curve modeling has been increasingly used in be havioral research including drug use in recent decade.'1 0 It provides a R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 7 way to model individual differences of an outcome variable. The underly ing process is conceptualized as continuous and expressed in latent growth factor means and variances. Different growth rates across differ ent groups can be compared. However, it cannot be used to identify sub groups with different growth trajectories and their covariates. Latent growth mixture modeling is a more general form of latent growth modeling with latent classes or mixtures.1 1 :, ! |S The growth factor means represent typical growth trajectories in a sample, and growth fac tor variances around these means represent heterogeneities in growth. This method enables researchers to identify discrete classes of individu als based on common patterns of growth. Risk factors or antecedents can be included in the model as covariates to examine their main, moderat ing. or mediating effects on the latent growth factors of adolescent smok ing. These analytical methods have been increasingly applied in research on drug use and abuse in recent decade. 2.7 Future Directions for Research on Adolescent Sm oking Based on the findings and the gaps in knowledge and methodology, the following future directions for research on the progression of adoles cent smoking are proposed: R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 8 1. Conduct studies to validate the measures of smoking stages by in cluding biochemical measurements: 2. Refine the definition of smoking stages and progressions with nec essary biological or clinical measurements: 3. Continue to identify factors or predictors that are most influential for individuals or groups at various stages in the smoking acquisi tion process: 4. Devote more effort to the study of the risk factors for higher levels of smoking and to identify the predictors of transitions to advanced stages of adolescent smoking by integrating biogenetic factors: 5. Determine the relative importance of intrapersonal, interpersonal, environmental influences in relation to the inheritance of biological factors: 6. Further identify associations between adolescent smoking and risk factors in a short interval when the risk factors are examined as time-invariant to minimize the possibility that the risk factors change over time: or examine these associations by modeling risk factors as time-variant variables: 7. Clarify the roles of parental attitudes and parenting styles on smoking initiation: R eproduced with perm ission of the copyright owner. Further reproduction prohibited without perm ission. 3 9 8. Determine how the impact of explicit pressure such as actual fam ily member’s or friends’ smoking compares with that of social per ceptions; 9. Examine the patterns and risk factors of adolescent smoking across gender, socioeconomic status, cultural background, and diverse ethnicities: 10. Conduct research to examine interactive and/or mediational effects among intrapersonal, interpersonal, and demographic risk factors on progressions of adolescent smoking; 11. Examine associations between developmental trend of social influ ences on long-term adolescent smoking behavior: 12. Use advanced statistical techniques such as latent growth mixture modeling to comprehensively investigate the developmental na ture. antecedents, and sequelae of adolescent smoking and its rela tions to other problem behaviors: 13.Tailor smoking prevention programs and interventions based on patterns of smoking progression, age. and ethnicity in addition to universal components; 14. Continue to clarify the content of prevention programs and the best time to implement intervention of smoking. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 4 0 1 CHAPTER 3 STUDY DESIGN AND MEASUREMENTS 3.1 Overview of Study Design The present study utilized the data from the project TOPP to con duct extended secondary data analyses. The ultimate objective of the present study was to further explore the longitudinal changes of adoles cent smoking and to examine associated risk factors. The most obvious advantage of longitudinal studies is its capability of examining change over time and identifying its determinants. The mathematical founda tions and statistical methods for analyzing longitudinal data have been proposed by Dwyer and Feinleib/’" ’ and Diggle, Liang, and Zeger.:,,i In the present research, new statistical methods for longitudinal data were ap plied to examine smoking progressions among adolescents. First, latent transition analysis (LTA)"’° was used to identify transition probability of adolescent smoking across stages, particularly progression towards addic tive smoking. Second, latent growth curve modeling (LGC),;- and latent growth mixture modeling (LGM)1|S were used to assess the longitudinal R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 41 patterns of growth trajectories of smoking in young adolescents (7th to 8th graders). Finally, generalized estimating equations (GEE) modeling1 -7 was used to assess the longitudinal effects of the project TOPP and the effects of demographic, intra- and inter-personal factors on regular and addictive smoking during adolescence adjusted for design effects. 3.2 Theoretical Model Based on prior studies and theories related to smoking progression, a hypothesized theoretical model for the current study was proposed (Figure 3.1). In this research, we hypothesized that progression of smok ing behavior (transition across stages or growth trajectories) could be di rectly predicted by intrapersonal characteristics such as attitudes, inten tions. self-efficacy, school performance, and interpersonal factors such as social learning (parents' and friends' smoking), social bonding (likeness of school), and social norm (prevalence estimates). Interpersonal and in trapersonal factors may interact with each other on the progression of adolescent smoking. The effects of interpersonal factors on smoking pro gression may also be mediated through exerting impact on the intraper sonal characteristics. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . Direct Effect M oderator Effect Prevention Program Demographic Characteristics: G ender, ethnicity, socioeconom ic statu s ■ Social learning * Social bonding • Social norm s Interpersonal Factors: Progression of Smoking Behavior: Growth curve/ trajectories Transition a c ro ss stag es • A ttitudes • Intentions • Self-efficacy • School performance Intrapersonal Factors: Figure 3.1 C onceptual m odel of psvcho-socio-behavioral influences on sm oking progression Prevention program and demographic characteristics may be di rectly associated with different patterns of smoking progression or they may moderate the effects of intrapersonal and/or interpersonal factors on the progression of adolescent smoking behavior. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm ission. 4 3 3.3 Hypotheses The following major hypotheses will be tested in the present study: (a) Patterns of smoking transitions across different smoking stages, especially progression to nicotine addiction, will vary by gender and ethnicity. We expect to identify 3 discrete longitudinal pat terns of trajectories of adolescent smoking based on smoking levels, the point at which smoking begins to escalate, and rate of escala tion. (b) Intrapersonal or psycho-behavioral variables will predict smoking progressions such that adolescents who have increased transitions or rapid growth trajectories of smoking will have higher intention to smoke, lower refusal self-efficacy for smoking, and lower aca demic grades compared to stable non-smokers. Intrapersonal vari ables will also mediate the effect of prevention program, interper sonal, or demographic variables on smoking progressions. (c) Interpersonal or social variables will predict different patterns of transitions or growth trajectories of adolescent smoking. Com pared to stable non-smokers, adolescents with quick transition to regular smoking or nicotine addiction or rapid escalation will have R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 44 more smoking friends and/or parents, higher level of perceived ap proval from their friends and/or parents, greater positive preva lence estimates, and lower school bonding. Interpersonal variables will interact with each other or with demographic variables such as gender or ethnicity on adolescent smoking progressions. (d) The tobacco prevention program will predict reduced smoking pro gressions. It will also influence the levels and growth rates of in trapersonal and interpersonal variables over time, which in turn will mediate the program effects on declined level and trajectories of adolescent smoking. 3.4 The Project TOPP The Tobacco Program and Policy Trial (TOPP) was a two-group randomized trial that followed 2,053 seventh-gi'ade adolescents over three time points during 1998 and 2000. The primary purpose of the project TOPP was to compare the effects of a multi-component school tobacco pol icy intervention with existing school policies on preventing tobacco use among middle school adolescents. The project was conducted in 19 Southern California schools in three school districts in Orange County. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 3.4.1 Intervention The prevention program consisted of four components: a 3-session school tobacco prevention curriculum, faculty presentation. Parent Teacher Association (PTA) presentation, and a faculty/administrative pol icy workshop. The school districts participated in this program were: Anaheim Union High School District (9 schools). Newport Mesa Unified School District (4 schools), and Irvine Unified School District (G schools). The student curriculum was composed of three 50-minute sessions, using the latest in tobacco prevention technology. The first session in cluded social influences and social norms. The second session focused on defining why we had policies and how these policies were beneficial to students. The third session focused on role-playing techniques to practice sharing the school's tobacco policies. A 10-to-lG-minute faculty presentation was made to all faculty, which focused on defining and clarifying each school’s policy and explain ing the importance of having, understanding, and disseminating the school policies. Included in this presentation was an explanation of the three types of school environments related to the consequences of tobacco use on campus (punishment, punishment with counseling, or counseling with support). R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 46 Parents were invited to attend a Parent Teacher Association (PTA) workshop that focused on parental-teen communications. Imbedded in this workshop was a discussion on their school’s tobacco policy. The par ents were also given information on the three types of school environ ments related to tobacco policy enforcement (punishment, punishment with counseling, or counseling with support). A one-hour workshop focusing on tobacco policy was conduced at each program school. Two administrators (usually the principal and the assistant principal), two teachers (usually the health teacher and the drug prevention coordinator), and two parents (usually the PTA president and a PTA member) were invited to attend the workshop. The focus of this workshop was to highlight the importance of creating, disseminating, and enforcing school tobacco policies. Each school was charged with ex amining its policies, given direction on how to evaluate its policies, and empowered to make changes if necessary. 3.4.2 Procedures The procedures involved three steps. First, an initial student self- report survey was distributed in each classroom by a pair of trained data collectors, which assessed the following: perceived social norms towards R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 47 tobacco use. social support for non-use among peers and school staff, and policy awareness, compliance, and support. Second, a three session to bacco health education program was implemented in half the 7th grade health classes as a supplement or component of the standard 6 hours of tobacco prevention education, while the other half received only the usual required six hours. The intervention was delivered with the trained health educators from the University of Southern California. Third, two post-test surveys (6 month and 18 month) were conducted to test for knowledge and behavioral change. 3.4.3 Data collection A consent procedure was used in which parents returned the con sent form only if they did not wish for their child to participate in the survey. This procedure was being used because the survey was confiden tial and did not assess students’ illicit behaviors, other behaviors that might be sensitive, or family life. The Institutional Review Board (IRB) of University of Southern California approved this variation of an active consent procedure for this evaluation. Trained data collectors collected data according to a standardized protocol. Data were batched by classroom period and school at three time R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 48 points. A baseline survey was conducted in the fall of 1998 when the sub ject started 7th grade. Two follow-up surveys were conducted in the spring of 1999 (6 months. 7th grade) and the spring of 2000 (18 months. 8th grade) respectively. Each survey took one class period (approximately 40 minutes) to complete. All information that was provided to us by stu dents. teachers, and principals were confidential. No report of the find ings at the individual (student) level was given. 3.5 Subjects The sample size for Project TOPP was large. A total of 2.G03 sev enth grade students across all selected schools completed the tobacco health survey at baseline (Time 1). giving a consent rate of 92.0% (eligi ble population was 2.830). The attrition rate was 9.0% at time 2 (n=2,40G). and 11.3% at time 3 (n = 2.135). The total sample size across all three waves was 2.053. giving a total completion rate of 78.9%, which served as the sample for the present study. Percentages of gender and ethnicity varied at each school and school district (Appendix A). Overall the longitudinal sample (n=2,053) was 48.5% males. 40.7% Whites, 2.4% African Americans. 29.9% His panic, 16.7% Asians, and 10.3% others. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 4 9 3.6 Measures Seven measures were conducted in the Project TOPP: ■ A student survey measured demographic data (gender and ethnic ity). intentions to use tobacco, tobacco use behaviors, perceptions of school tobacco policies, and perceived norms and consequences of tobacco use (n=2.0;)3). ■ A principal policy interview assessed variation in levels and types of policy enforcement, perceived tobacco use problems in and around school, and changes in policy during the course of the study (n=19). • A tobacco policy survey, completed by on PTA member, one teacher, and one staff administrator per school (n=19). assessed school tobacco policy review and change: student participation in tobacco programs: tobacco policy dissemination, awareness, imple mentation. and support: and perceived student tobacco use. ■ A school tobacco program matrix, completed by one school staff administrator per school (n=19). assessed tobacco program, courses, assemblies, bulletins/newsletters, counseling, and parent information, which was implemented during the school year. R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 5 0 ■ Written copies of school district policies were collected for qualita tive analysis. ■ Written copies of school policies were collected for qualitative analysis: ■ School records of tobacco and other drug-related infractions were collected. Each of these measures was collected annually, with the exception of the student survey, which was colleted at baseline. G months, and 18 months. In the present research, only student data were analyzed. The original student survey of Project TOPP contained 54 items, which in cluded demographic variables, perceived social norms, perceived conse quences of tobacco use and policy violation, social support for non-use from parents and peers, and frequency and amount of tobacco use. All of these measures have been used before, with acceptable reliability and va lidity (r> ranging from .GO to .97. sec Table 3.1).H i’ ’ U i~ The following were the measured items that were implemented in the Project TOPP and were analyzed in the present research (refer to Table 3.1 or Appendix B for detailed descriptions and scales of these measures). R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 3.6.1 Outcome measures 3.6.1.1 Cigarette use variables: ■ Number of cigarettes smoked in one’s whole life ■ Number of cigarettes smoked in the past 30 days ■ Number of cigarettes smoked per day ■ Use of tobacco within the first 30 minutes of waking 3.6.1.2 Defin ition-based smoking stages: • Abstainers were defined as those who never smoked in their whole life. • Triers were defined as those who smoked part or a whole cigarette in their whole life. • Experimenters were defined as those who smoked more than one but less than 100 cigarettes in their whole life, and did not smoke in the past 30 days. ■ Current smokers were defined as those who smoked more than one but less than 100 cigarettes in their whole life, and smoked in the past 30 days. ■ Regular or established smokers were defined as those who smoked more than 100 cigarettes in their whole life and R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . smoked cigarettes in the past 30 days, or smoked more than 10 cigarettes per day. or used tobacco within 30 minutes of waking. ■ Addictive smokers were defined based on simplified Fager- strom Test for Nicotine Dependence (FTND) score (greater or equal to 1). which were obtained by summing the scores for the follow two items: use tobacco within the first 30 min utes of waking (no=0. yes=2). and number of cigarettes smoked per day: (<10 cigs=0. 11-20 cigs=l. >20 cigs=2). ■ Quitters were defined as those who were regular smokers, but they did not smoke in the past 30 days. For longitudinal data analyses in this study, triers, experimenters, and quitters were re-categorized together as trier/experimenters or non- current smokers because triers and experimenters had similar features in the extent of smoking and there was no enough validating information to differentiate quitters from experimenters in this research. Thus, in latent growth mixture models, the values of smoking stages were: ^n o n - smoker. 2=non-current smoker, 3=monthly smoker. -l=daily smoker, and o=addictive smoker. R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 53 3.6.2 Predictors 3.6.2.1 Interpersonal or social factors ■ Number of parents who smoke ■ Number of friends who smoke ■ Social norms 3.6.2.2 Intrapersonal or psycho-behavioral factors ■ Intention to smoke ■ Refusal self-efficacy ■ Academic grades (GPA) 3.6.2.3 Demographic characteristics and prevention groups ■ Gender ■ Ethnicity ■ School-level social-economic status (SES) ■ Prevention groups 3.6.3 Reliability estimates of the measured variables The smoking items in the present study were adopted from the pro ject SMART.1 < if) The reliabilities of smoking variables, intrapersonal, and R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 5 4 interpersonal variables used in this research were high based on previous studies and in our data as well (see Table 3.1).1 0 1 s _ > -1 7 :5 ls0 The reliability of lifetime cigarette use (r = .90) was higher than monthly cigarette use (r - .64) and daily cigarette use (r = .60).s r> Out of the six items of Fagerstron Test for Nicotine Dependence (FTND). the number of cigarettes smoked per day (r = .97) and time to the first ciga rette (r = .82) had the highest reliabilities.17* Intention to smoke (r - .74) and refusal self-efficacy (r = .66) had moderate test-retest reliability.s r> Adolescent self-reported parental smoking was shown to be a reliable measure of parental actual smoking (Kappa = .82 for mothers and .72 for fathers).1 0 Adolescent self-reported friends' smoking was also reliable over time (Cronbach coefficient a = .71).,s0 Self-reported academic grades were found to be highly correlated with school archival information of academic grades (r = .78).7 ,t The internal consistency as measured by Cronbach coefficient al pha for most measured variable over three time points in the present re search was high (range from .67 to .92) (see Table 3.1). However, for monthly, daily, and addictive smoking measures, the internal consistency was relatively low (range from .40 to .58), indicating the changing nature of these variables over time. R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . oo Table 3.1 R eliability estim ates of the m easu red variables Variable M easures and scales C ronbach A lphar Test-retest R eliabilitv: 1 0 .7:5 .90 s' .61 .60 1 How many cigarettes have you smoked Lifetime in your whole life? ( l-nonc, 2=part or cigarette use all one rig., 3=2-10 cigs., -1=11-20 cigs, .92 5=1-5 packs, f>=niorc than 5 packs.) 2. How many cigarettes have you smoked Monthly in the past 30 days.’ ( l=nonc, 2=part or cigarette use all one cig., 3=2-10 cigs., 1=11-20 cigs, .68 5=1-5 packs, 6=morc than 5 packs.) 3. During the past month, on the days you smoked, how many cigarettes did you Daily ciga- smoked per day? ( l=did not smoked, .66 rette use 2=lcss than I cig., 3=1 cig, -1=2-5 cigs. 5=6-10 cigs, 6=11-20 cigs, T=rnore than 20 cigs.) 1. Do you use tobacco within the first 30 Addictive minutes of waking'.’ (l=no, 2=yes, .67 smoking 3=don’ t smoke) 6. Smoke 10 cigarettes per day. (0=no, l=yes) 6. Do you think you will smoke a cigarette Intention to in the next few months'.’ ( l=definitcly smoke not, 2=not sure, 3=proh<ibly, -l=def’ initcly yes) 7. If your best friend offered you ciga- Refusal self- rettes. how hard would it be to refuse .67 efficacy the offer? ( I =i cry hard. 2=hard, 3=ecisy, ■l=i cry easy) 8. Of the two adults who are the most mi- Parental portant in your life, how many do you .86 smoking think use cigarettes? <l=none, 2=1 adult, 3=2 adults) Friend 9. Of your five friends, how many smoke smoking cigarettes? <.1=0, 2=1, 3=2, -1=3, 5=1, .70 6=5) 10. What grades do you usually get in Academic school? <I=mostly ,-1's, 2=A’ s an d B ’ s, grades 3=mostly B's, -l=B's an d C's, 5=Mostlv .S-l (GPA) C ’ s, 6=("s and D ’ s, 7=Mostly D ’ s, S=D’ s a n d F's, 9=mostly F ’ s) .82 171 .97 171 .71 .66 v " ' 93.1% (mother). 86.-1";, (fa ther) 10 * .71 .78 ‘ N ote. t. S ta n d a rd ized C ronbach alp h a w a s calcu la ted based on the m ea su red v a r i ables across th ree w a v e s in th e presen t stu d y . ±. T est-retest correlation w a s cited based on past stu d ie s w ith sim ila r m ea su re s. §. Exact a g r e e m e n t w a s sh o w n b etw een a d o le sce n t self-report a n d m other's and father's a ctu a l sm ok in g. C orrelation coefficient w a s sh o w n b e tw e e n ad o lescen t self-report and school archival inform ation. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth er reproduction prohibited w ithout perm issio n . 1 CHAPTER 4 5 6 NEW STATISTICAL METHODS FOR ASSESSING CHANGE It is critically important to measure the individual change (pro gression, transition, or growth trajectory) of adolescent smoking longitu dinally because by measuring change over time accurately we can map out critical patterns and associated time-variant or time-stable predictors during adolescent development. The mathematical foundations and sta tistical methods for analyzing longitudinal data have been proposed by Dwyer and Feinlcib.'’’ ’ and Diggle. Liang, and Zeger.’ ,(i Conventional sta tistical methods for measuring the changes of individuals have primarily focused on the analysis of change between two waves (pretest-posttest change scores) or the analysis of change in one variable at a time (uni variate change). However, when data involve multiple time points (greater than two) or involve multiple outcome variables, more advanced statistical methods are needed.212 Structural equation modeling (SEM) has been widely used in psy chological and social science, and increasingly used in medical research to R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . evaluate causal hypotheses. Structural equation modeling is a powerful statistical method to account for measurement errors and examine the causal relations between latent constructs and their measured indicators and between latent constructs in both cross-sectional and longitudinal designs.1 !-l,1(ls Numerous longitudinal studies have been conducted us ing a latent variable SEM approach to identify potential causal links be tween risk or protective factors and adolescent substance use and abuse.1 lir’11‘ )i 1 ,1 1 Mediator and/or moderator models have also been tested within a latent variable SEM framework.1 - 11')11 ,1 1 Although conventional SEM has capabilities to examine longitudi nal changes of outcome variables or latent constructs by controlling base line measures in longitudinal studies.1'’- 1 (1 1 it does not enable modeling individual’s developmental growth rates or trajectories over time. Con structed within a latent variable SEM framework, however, the latent growth curve (LGC) modeling has been developed to capture individual’s developmental trajectories and variations in these trajectories over time.'*2 1 ir> 1 1 (1 Latent growth curve modeling provides a mean of modeling development as a factor of repeated observations over time. Within the LGC framework, age or time is viewed as a dimension along which be havior changes (e.g. smoking progression) form the dependent variable in developmental studies. LGC is concerned with the study of individual R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 5 8 differences in development over time, which is typically captured by ran dom effects. SEM is concerned with relations among observed and latent variables. LGC and SEM, on the other hand, are connected because ran dom coefficients (i.e., growth parameters) may be modeled as continuous latent variables.11 0 Thus LGC benefits from the full generality of a latent variable SEM framework.11 7 Because LGC goes beyond conventional SEM and possesses advantages of modeling individual developmental growth rates and their variations over time, it has been considered as second-generation structural equation modeling.11 ,1 Recent years have witnessed major advances in the statistical methods available for more advanced quantitative analysis of longitudi nal data. In particular, new advanced statistical methods such as latent transition analysis (LTA). latent growth curve (LGC) modeling, latent growth mixture modeling (LGM), mixed-effects (random-effeets. multi level) modeling for both continuous and dichotomous longitudinal data, and generalized estimating equations (GEE) modeling have been devel oped and applied in the arena of behavioral research in recent several years. In the following sections, the brief technical perspectives of the advanced statistical methods are provided followed by the detailed proce dures of their applications in the present research. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 4.1 Latent Transition Analysis (LTA) 5 9 Latent transition analysis (LTA) expands the latent Markov model to allow applications to more complex latent variables and the use of multiple indicators.1 8 -™ LTA is rooted in latent class theory which is a measurement theory based on the idea of a discontinuous latent variable that divides a population into various mutually exclusive latent classes.s- LTA has been used to test stage-sequential models and it has been ap plied in health behavioral research such as substance use prevention.88 alcohol abuse.8 ,1 smoking uptake."'118,i smoking cessation.1 81 and other health-risk behaviors.1 "7 The main objective using LTA in our study was to identify smoking stages and to examine the transition probabilities of smoking stages across three time points. 4.1.1 Overview of smoking transition models As shown in Figure 4.1, five latent status (stages) of smoking was proposed. Non-smokers may progress following the sequence of non- smokers, trier/experimenters, current smokers, regular smokers, and fi nally. addictive smokers. It is also allowed that transition skips stages without following the continuous sequence. For example, non-smokers R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . G O may directly progress to current smokers, or even addictive smokers, al though the probability is small. The same is true for the progression be tween other stages. B a c k w a r d m o v e m e n t Triers / Experi menters Addictive smokers n n U Figure 4.1 H ypothesized model for sm oking tra n sitio n across stages On the other hand, backward progression or regression from higher stages to lower stages is also allowed in the LTA model with an R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . G1 exception for regression to non-smokers because once an individual be comes a smoker, he/she could be either a current smoker or a ex-smoker or non-current smoker, but could never be a non-smoker again. For ex ample. addictive smokers may reduce their smoking stages to regular smokers, or to current smokers, or trier/experimenters (non-current smokers), but not move backward to non-smokers. 4.1.2 Parameter estimates in LTA models Three sets of parameters are estimated in an LTA model by means of the EM algorithm.vt which include measurement parameters (p pa rameters), unconditional probabilities of latent status membership (cr pa rameters), and transition probabilities ( r parameters). All parameter es timates were calculated using the WinLTA program (free download avail able at http://methcenter.psu.edu).I(i The measurement parameters (p parameters) represent the prob abilities that an individual responds “yesT ’ to a particular item, condition ing on membership in a particular latent status. The estimated p pa rameters serve as two roles: First they define the latent status: second they reflect measurement reliability or precision. In this research, four measured variables (i.e., lifetime smoking, monthly smoking, daily smok R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . ing, and FTND scores) were used to define five latent statuses: non- smokers, trier/experimenters (non-current smokers), current smokers, regular, and addictive smokers (see Figure 4.1). By definition, non- smokers are those who answered "No" for all four measured variables. Thus the estimated p parameters for the "non-smokers" latent status should be low on all four measured variables. In contrast, the estimated parameters for "addictive smokers" latent status should he high on all four measured variables. The unconditional probabilities of latent status membership (<r pa rameters) represent the unconditional probabilities of membership in each latent status at each time point. They reflect the proportion of indi viduals in each latent status out of entire sample. The transition probability matrix ( r parameters) is the heart of the LTA analysis. They reflect the transition probabilities between latent statuses (smoking stages) from one time point to another, conditional on latent class membership. In our study, one transition probability is the probability of moving to other smoking stages at time 2, given member ship in each smoking stage at time 1. The following is the representation of a transition probability matrix for five latent statuses between time 1 and time 2: R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 63 Time 1 Tim e 2 ■ i i ■1 4 •I 5 r-, r i ? r . r, > i •4 I r>; Ti, r5, r24 ^4 ^44 ^4 f*2 ^ ^ ^ ^*4 ^ ^ ^ In this matrix. r„ h denotes the probability of membership in latent status b at time 2. conditional on membership in latent status a at time 1. In our study, for example, r/ _ • represents the probability of member ship in triers/experimenters latent status at time 2. given the non- smokers latent status at time 1. It may interpret as the probability of an adolescent in the "triers/experimenters” latent status (latent status 2) at time 2 if he/she starts out in the "non-smokers" latent status (latent status 1) at time 1. The total probability in a row sums up to one because they are conditional probabilities. 4.1.3 Model fit index in LTA models In LTA models, the goodness Gf fit is assessed by comparing the observed and predicted response pattern frequencies. If the hypothesized R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n. G4 model provides a good representation of the data, then the predicted fre quencies will be close to the observed frequencies. In contrast, a poor model will not reproduce the observed response pattern frequencies ade quately. The Win LTA program output provides the likelihood-ratio chi- square. usually denoted G~. This quantity is computed as follows: C; = 2 I Where: f lk represents the observed frequency of response pattern ijk, and f k represents the frequency predicted by the model. Asymptotically. C 7" is distributed as a chi-square with degrees of freedom equal to K-P-l, where K is the number of possible response pat terns (i.e.. the size of the contingency table: maximum possible number of response patterns: K = 2". n is the total number of dichotomized observed variables in LTA models) and P is the number of parameters estimated. When the data are sparse (i.e.. there are few or no observations in many cells), the chi-square distribution is usually not a good approximation for the distribution of G: . Sparse tables are likely to be a problem whenever complex LTA models are estimated. As a partial remedy for the problems associated with the distribu tion of G~, a double cross-validation1 7 was conducted. Double cross- validation involves splitting a sample into two (or more) sub-samples, for R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . example. Sample A ancl Sample B. and fitting a series of plausible models to each sample. Each model is fitted to Sample A (the calibration sam ple). the predicted response frequencies for each model are compared to the observed response frequencies in Sample B (the cross-validation sam ple), and G~ is computed. Then the reverse is done: each model is fitted to Sample B (now the calibration sample), the predicted response fre quencies for this model are compared to the observed response frequen cies in Sample A (now the cross-validation sample), and another G~ is computed. A model cross-validates well if the G~ is relatively small when the estimated model is applied to a cross-validation sample. When a series of models are tested, the model or models that cross-validate best are considered best-fitting. 4.1.4 Sign ificance tests for param eter estimates Although theoretically possible for the significance tests of pa rameter estimates in LTA models1 7 '’ 177, the current statistical pro gram, WinLTA (version 2.3). is not able to calculate the standard er rors of the parameter estimates and conduct significance tests for spe cific parameters across groups. For the omnibus tests of overall pat terns of unconditional and transition probabilities of latent status, a R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . GG chi-square test on the change of G~ and degree of freedom between two nested models, i.e.. a basic model with all parameter to be freely esti mated. and an alternative model with the unconditional or transition probability estimates constrained to be equal between two groups was performed. To compare certain specific transition probabilities, an al ternative R x C contingency table .f- test approach*1 was used based on the predicted marginal numbers of membership. 4.2 Latent Growth Mixture Modeling (LGM) Latent growth modeling has become a common way to model indi vidual differences in growth in substance use research.1 ’ ’ The conven tional growth model allows heterogeneity corresponding to different growth trajectories across individuals; however, it cannot capture hetero geneity that corresponds to qualitatively different development. Latent growth mixture modeling is a more general form of latent growth model ing with latent classes or mixtures.1 1 8 The means of growth factors rep resent the typical growth trajectory in the sample, and variance around these means represents heterogeneity in growth. The goal in the present research was to identify discrete classes of individuals based on the com mon patterns of adolescent smoking growth trajectories. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 6 7 4.2.1 Overview of Latent Growth Mixture Modeling Muthen and colleagues11 1 1 "il 1 S generalized the latent growth curve (LGC) modeling to latent mixture trajectory models and proposed a framework for general growth mixture modeling (GGMiM). The GGMM approach extends the conventional growth model by modeling heteroge neity in the population and examining simultaneously the influences of different growth trajectories, predictors (background variables), concur rent process variables, and the sequelae of the latent class variables. La tent growth mixture models handle heterogeneity in growth trajectories by identifying discrete classes or mixture. The censored normal (CNORM) model was used in the present study because it was useful for modeling the conditional distribution of psychometric scale data, given group membership.1 1 ,7 l',,l A distribution allowing for censoring was used because the smoking statuses in our data tended to cluster at the minimum {Min) or the maximum (Max) of the scale . The likelihood of observing the data trajectory for subject i. given the group membership k, is P r ( >; = y C =k.l ( ; = » ■ ) = * Min-.u,,. n 1 1 __ i n i XT a \ftn i ,U c v O ’ a v : Mu.\ <y R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 6 8 Whore /'„* = ft* + age,, ft, t + agetl 2 ftz k + ■ • ■ + n -,,ft (4.2) The censored normal model is also appropriate for continuous data with approximately normal distribution, with or without censoring. The uncensored normal model is applied by specifying minimum and maxi mum values that lie beyond the range of the observed data. Subjects with missing longitudinal data or time-variant covariate values are well handled in latent growth mixture models by maximum likelihood estimating methods. However, subjects with any missing risk factor (time-invariant) values are excluded from the analysis. 1.2.2 Model selection The number of latent classes and the order (linear, quadratic, or cubic) of the model were assessed empirically by the Bayesian Informa tion Criterion (BIC) or Akaike Information Criterion (AIC). The model with smallest BIC or AIC is favored. The BIC1S I is defined as B IC - -21og£ -win//. (4.3) Where logL is the log likelihood of the model, r is the number of free model parameters, and n is the sample size. R e p ro d u c e d with p e rm ission of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm ission. 6 9 The AIC! is defined as .-1/C ' — — 2 log L + 2/\ (4.4) As suggested by Jones and colleagues.10’ the change of BIC (.ABIC) was used to assess the degree of evidence favoring the alternative model (e.g.. .ABIC = 6-10 as strong evidence against Ho. ABIC > 10 as very strong evidence against Ho). 4.2.3 Modeling strategy The goal of this research was to identify classes of adolescents on the basis of the trajectories of smoking status over time (1 = nonsmoker. 2 = triers/ experimenters. 3 - current smokers, 4 = regular smokers. 5 = addictive smokers) defined by four measured variables (lifetime, monthly, daily, and addictive use) as conceptualized in Figure 4.2. In the concep tual latent growth mixture model, the following estimations were allowed jointly: (a) different growth trajectories of a finite growth mixture model captured by class-varying random coefficient means and variances; (b) varying class membership probabilities and associated antecedents: and (c) log odds of antecedents on latent class members using a logistic re gression model. Linear model with one latent class was modeled as an initial step, and models with more latent classes and higher orders were R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth er reproduction prohibited w ithout perm issio n . 7 0 examined in sequence. The final quadratic mixture model with three la tent classes was selected based on the smallest BIC and AIC values. The models without any covariates were examined to obtain proportion of la tent class membership. Finally predictors were added in the model to es timate the relative risk of higher level of latent classes in comparison with lowest level of latent classes. Growth Factors Trajectory Classes Tobacco Use over Time Consequences: Nicotine Addiction Antecedents: Interpersonal Intrapersonal Demographic Intervention Figure 4.2 Conceptual latent grow th mixture model of smoking. R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 4.3 Mediator and Moderator Models 7 1 As defined by Baron and Kenny11, a mediator variable is "the gen eral mechanism through which the focal independent variable is able to influence the dependent variable of interest (pp. 1173)." A mediating ef fect implies that the independent variable influences the mediator that, in turn, influences the dependent variable. To establish a mediating ef fect as shown in Figure 4.3 A. three criteria need to be met. First, inde pendent variable must influence the mediator: second, independent vari able muse influence the dependent variable: third, the mediator must in fluence the dependent variable and the influence of the independent variable on the dependent variable is reduced after controlling for the mediator.11 A moderating effect is distinct from a mediating effect conceptually and methodologically. According to Baron and Kenny11, a moderated or moderator effect refers to "an interaction between a focal independent variable and a factor that specifies the appropriate conditions for its op eration (pp. 1174)." In other words, a moderator variable influences the relation between an independent variable and a dependent variable in such a way that the effect of the independent variable on the dependent variable varies based on the levels of the moderator variable. The form of R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n. this interaction is analogous to that of the buffer interaction effect pro posed by Cohen and Wills.1 1 To establish a moderated effect as shown in Figure 4.3 B. the interaction (y) between the independent variable (X) and the moderator (Mo) on dependent variable (Y) should be significant.1 1 As indicated by Holmbeck!,s. there were large inconsistencies in the use of the terms ■'mediator" and "moderator" in psychological and behavioral studies conceptually and statistically. In addition to the distinctions in above-stated definitions and sta tistical methods between mediators and moderators, temporal sequence among independent variable, dependent variable, and a third variable is also key element to distinct a mediator from a moderator. Moderation does not assume connotation of causality, while mediation implies at the minimum a causal order, and often additional causal implications are re quired to explain how mediation occurred. Thus, it is purposeful in the use of the terms independent and dependent in moderator models, and antecedent and consequence in mediator models. However, things are not necessarily as straightforward as the aforementioned distinctions, one reason being that mediating variables may involve or he influenced by a moderator creating a situation termed "moderated mediation."ll)- R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm ission. 7 3 Figure 4.3 Me W here: \ = Independent variable: Me = Mediator; Y = Dependent variable. A. Hypothesized M ediational Model Mo X*Mo W here: X = Independent variable: Mo = M oderator: Y = Dependent variable. B. Hypothesized M od erato r Model Theoretical models of m ediation (A) and m oderation (B). R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 7 4 In some circumstances, both mediated and moderated effects can be tested within one research. Some variables may serve as both media tor and moderator on behavioral development process. However, under any circumstances, terms, conceptions, and statistical methods used to test a mediator or moderator effect in a study should be consistent. 4.4 Mediating Process of Latent Growth Curve Modeling Intrapersonal variables such as intention to smoke and refusal self-efficacy could be modeled as mediators. Mediational analysis has been applied in estimating intervention effects in longitudinal stud ies.11 ,1 1!()l 1 1 1 ,1 ,1 1 ,i'’ Three separate models were tested. Model 1 assessed the effects of the prevention program on intrapersonal factors (mediator). Model 2 assessed the effects of the prevention program on growth factors of lifetime smoking. Model 3 included both mediator and outcome vari able to assess the effects of the prevention program on outcome control ling for mediator. The detailed procedures suggested by Baron and Kenny1 1 were followed to estimate the total, direct, and mediated effects and to test the statistical significance. In the framework of latent growth curve modeling (LGC). the mediational analysis was conducted in a fashion of two distinct parallel R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . processes.11 7 The objective was to examine the growth trajectory for both the outcome variables and the mediators simultaneously and to identify the mediational effects by examining the relations among independent variable (e.g.. the prevention program), the latent growth factors of the mediators and the outcome variables in the same manner discussed above. The hypothesized mediational model in LGC framework was pro posed (see Figure 5.7). in which a two-level mediation was assumed be cause the prevention program was randomized and administered at school level and the intention to smoke (mediator) and lifetime smoking (outcome) variables were assessed at individual level.1 1‘* The following illustrates the analysis for testing the effects of the tobacco prevention program (independent variable) on growth rates of adolescent tobacco use (dependent variable) mediated through the growth rates of intention to use tobacco (mediator). (a) Dependent variable process: Ylt = Intiv.i + Slp.vi) * t + r.(Y iu (4.5) Int,Yii = IntiYou + 7 1 * X, + '; ■ > * Int,Mu + v intiY o (4.G) Sip,YD = Sip,Y(h) + T * X, + V ( * IntiM,) + P * Slp,M„ + v siptY n (4.7) (b) Mediator variable process: M,t = IntiM,) + Sip,Mi) * t + EiM .n (4.8) R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth er reproduction prohibited w ithout perm issio n . 7 6 (4.9) SlpiMo — SlpiMOi) " F a X, F - V'SlpiMn (4.10) Where Y„ is the dependent variable (e.g.. lifetime smoking) at time t: IntiYn and Int,M ,> are the initial status of the dependent variable and the mediator variable (e.g.. intention to smoke): Slp(Y o and Slpcu,) are the slopes of the dependent and mediator variables: Inhvon. InhMo,). Slpivon. and SlpiM Dn are the intercepts of the growth parameters: X, is the inde pendent variable. Mediated effect can be estimated by the following for- mulas:>"'' i’1 1 i‘» - Mediated effect = (/.x[) (-1.11) Standard error of t/xf): < > ,. , = ■ + p :S'~ + (4.12) 1.5 Interaction between Latent Growth Factors Moderating (moderator, interactive, or buffering) effects between demographic, social influence, and intrapersonal factors on adolescent smoking behavior can he examined using interactive models.,,(i Statisti cal analysis for moderation was conducted by including an third term that was multiplied by the two predictors in multivariate logistic regres (4.13) R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . sion models and the examination of interactive effect will he guided by the procedure outlined in Baron and Kenny." and Dwyer.'" In the latent growth curve framework, the interactions among growth parameters were examined based upon the models proposed in structural equation modeling framework.um 1 1 7 and extended to latent growth curve model ing. 1 2 :1 A comprehensive illustration of these models can be found in an introductory book on latent growth curve modeling.'’- The hypothesized moderator model in the LGC framework was proposed and illustrated in Figure 5.8. In multivariate regression models, to test the interaction between parental smoking and friends' smoking on the latent classes of growth trajectories, we first included parental and friends’ smoking variables in the model to test the main effects: then added a third variable by multi plying the prior two variables. The moderated effect was examined by the Wald test for the regression coefficients of the interaction term in the model. In the latent growth curve framework, a hypothesized model test ing the interaction between friends’ smoking and refusal self-efficacy on adolescent lifetime smoking progression is illustrated in Figure 5.8. In this model. Xi, is friends’ smoking at time i: Xu, is refusal self-efficacy at time i: Xt,X:>, is the interaction between Xi and \> at time i: Tob, is the R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . lifetime smoking at time i: rpi, r|2i, r|nr|2t, and the level of tobacco are the latent intercept factors: r|i2. r|2 2. r|i2ii22, and the trend of tobacco are the latent slope factors. The moderator effects of friends' smoking and re fusal self-efficacy on smoking progression were examined by testing the coefficients of the paths from r|nr| 2 i to the level of tobacco and the path from r| 121122 to the trend of tobacco. 4.6 Model fit indices for latent growth curve modeling A number of model fit indices are available in Mplus program (pp 3of)>.1 In addition to BIC and AIC. the following two model fit indi ces were used in the present research: (a) Comparative Fit Indices (CFI): In formula 4.15. e =/;(/? + I) / 2, p is the number of outcome vari ables. and r. is a residual in a correlation metric CFI = 1 - ma.\( x,, ' - d n .0) max( ~ ~ d„ . x,< ~ ~ > • fk 14) (b) Standardized Root Mean Square Residual (SRMR): SRMR = IT V \7 r, (4.15) r (4.16) R e p ro d u c e d with p e rm ission of th e copyright ow ner. F u rth er reproduction prohibited w ithout perm ission. 7 9 where s,t and cx,..are the sample and model-estimated covariance be tween the continuous outcome variables y and r; . Hu and Bentler1 00 suggests the following fit index cut off value guide for good models with continuous outcome variables: CFI > .95. SRMR < .08. Models with overall fit index of greater than .9 are empiri cally acceptable for a good model.1 7 4.7 Generalized Estimating Equations (GEE) Modeling In Project TOPP. sampling randomization was carried out at the school level (n=19 schools), and subjects (n=2.053 individuals) were clus tered in 19 schools (mean cluster size=l 15.07). Furthermore, each indi vidual was longitudinally surveyed at three time points (baseline, 6 months, and 18 months). Most standard statistical methods including ordinary logistic regression models assume that each observation that makes up a data set is independent of all of the others.7 1 However, in clustered randomization or repeated measurement designs, this assump tion is inappropriate because observations within each cluster (clustered design) or within an individual (repeated measurement design) tend to be correlated with one another.7 1 M I 1 2 1 Unless the within-cluster or with-in individual correlation that results from the sampling design is R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 8 0 accounted for in statistical analyses, the standard errors of the parameter estimates are biased, usually towards inflating the parameter’s statistical significance. 4.7.1 Overview of generalized estimating equations Generalized estimating equations (GEE) modeling is a simple ap proach to account for within cluster or individual correlations.1 -7 -1 " The GEE extends generalized linear models two ways. First, a GEE model allows the correlation of outcomes within an individual to he estimated and taken into account in the formulae which generate the regression co efficients and their standard errors. Second. GEE models allow the calcu lation of robust estimates for the standard errors of the regression coeffi cients.1 -7 Compared to ordinary logistic regression models which basi cally ignore the correlation within clusters, the GEE methods yield as ymptotically unbiased standard errors corrected for with-in cluster corre lations. Compared to the statistical methods of aggregated level data,i(i U i') that have been generally used to evaluate school-based prevention ef fects. the GEE method is more flexible because it applies to nearly any kind of dependent variables, including binary, categorical, continuous, and time-to-event data, and it keeps original measures intact. Compared R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 81 to random-effects modeling which is used most often for data with re peated measures or longitudinal data over time21 !,r> 121, the GEE method is used most often to analyze cluster-correlated data, particularly when the cluster size is large.n,i 4.7.2 Steps of GEE modeling Fitting a GEE model is conveniently viewed as the following steps: i. Fit a standard (naive) regression model by assuming all observa tions to be independent. In the current research, an ordinary logis tic regression (OLR) model was chosen for binary data. The log- odds of the expected value of the response is model using the fol lowing formula: log - v. = x„p. (4.17) Where r.( is a binary outcome (e.g.. regular smoking), taking on the value of 0 (not regular smoker) and 1 (regular smoker), .v ,, is a vec tor of covariates. ii. Choose the form of a working correlation matrix to describe the within-cluster correlation. Four types of correlation structures are R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 8 2 assumed in GEE modeling, i.e.. independent, exchange or com pound symmetry, autoregressive, and unstructured. iii. The parameter estimates of GEE models are the solution of the generalized estimating equations: £''(/?) = X '( F -//,), U-18) , -i rp where ut = (//,.•••.//„,. ) is the mean response vector for responses ( - 'L i i within a cluster. ------ is the vector of first partial derivatives of the eft mean response //( with respect to the regression coefficients fi F ’1 is the covariance matrix, which is inserted to weight the data. iv. A robust variance estimate for the estimated parameters is pro posed1 -7 as follows: l'(fj) = A/„ ' M] A/,, (4.19) where A/n = f (IL y; 1 (4.20) rP rP = - m , <4.2i) i - 1 ( p r p This robust variance estimate consists of three parts and is there fore termed a sandwich estimator. Under cluster sampling or repeated R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth er reproduction prohibited w ithout perm issio n . 8 3 measure design, the inclusion of M, serves as a variance correction when the correlation model is misspecified. 4.8 Intraclass Correlation (ICC) and Design Effects The intraclass correlation (p) is a measure of the degree of similar ity among individuals who belong to the same cluster or group, or among repeated observations over time within an individual. The intraclass cor relation (ICC) for continuous variables can be estimated by a standard one-way analysis of variance (ANOYA). in which the sampling unit such as school is the only grouping factor.11- or a two-level random intercept model.-1 The ICC for binary variables can be estimated in a random- intercepts logistic model.!,i l!)- The following are the formulas to calculate ICC for continuous and binary variables: P r continuous 0 (4.22) or binan (4.23) In Equation 4.22. r„~ is the estimated variance of a random inter cept assuming normal distribution between schools, and a 2 is the aver R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 8 4 age random residual variance within schools. In Equation 4.23. cr ' is the estimated variance of a random intercept assuming normal distribution. i / and/T‘/3 is the variance of a standard logistic distribution. The design effect is a useful analytic method because its size de termines the direction and extent of the bias imposed by the assumption of no vvithin-cluster correlation.1 ’ ’1 ’ The magnitude of the design effect is interpreted as increase (or decrease) in the variance accounting for intra class correlation compared to the variance under the assumption of inde pendence. The design effect is more meaningful than ICC alone because it is sensitive to the average cluster size as well as the ICC. It can be es timated by the following formula: Design effect = 1 + (m - 1 )p, (4.24) where m is the cluster size, and p is the intraclass correlation coefficient. For the case of unequal cluster size, the average cluster size is calculated by the following formula:lfS ’ ’ y m : Jil = J=-J——, (4.25) I , » ‘ , where m fs the number of observations in the /th cluster (e.g.. a school). R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 4.9 Statistical Packages The PROC TRAJ procedure1 '1 7 of the SAS statistical package18- was used to conduct latent growth mixture modeling. The results were vali dated and compared using Mplus program.1 1 1 1 The WinLTA program (version 2.3 for windows)1 * ’ was used to conduct latent transition analysis. Mplus version 2.021 1 1 1 was used to conduct multilevel growth curve analy sis and tests for interactions between latent growth factors. Intraclass correlation (ICC) for continuous dependent variables was estimated using the SAS PROC MIXED procedure1 '1 0 and validated by the Mplus pro gram.1 1 1 1 The ICC for dichotomized variables was estimated using ran- dom-effects logistic regression model of the SAS PROC NLMIXED18- and validated using MIXOR program.'1 1 Generalized estimating equations (GEE) modeling was conducted using the SAS PROC GEXMOD proce dure.18- All other statistical analyses such as prevalence rates, attrition analyses, ordinal logistic regression analyses, and descriptive and uni variate analyses were conducted using SAS statistical package.1 8- The SAS codes, latent transition analysis programs and parameter estimates, and the Mplus scripts are provided in Appendix C, D. and E. R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 8 6 ________________________________________________________________________________ c CHAPTER 5 RESULTS 5.1 General Characteristics of Adolescent Smoking To examine the overall characteristics of prevalence rates, smoking stages, and patterns of regular and addictive smoking, descriptive and univariate statistical analyses were performed. The results included attrition analysis, comparisons of prevalence rates between school districts, gender, ethnicity, and prevention groups, overall patterns of smoking stages, and patterns of regular and addictive smoking over time. 5.1.1 Attrition Analysis The overall attrition rate of our data was 21.1% across three time points. African Americans, Hispanics, American Indians, adolescents in low socio-economic status, and smoked cigarettes at baseline were more likely to drop out both from time 1 to time 2 and from time 2 to time 3 (Table 5.1). R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 8 7 Table 5.1 Attrition rates at time 2 and time 3 by demographic char acteristics and smoking status at baseline Character From T 1 to T2 From T2 to Td Total (from T1 to Td) Stayer X IM Dropout X Co) Stayer X Co) Dropout X Co) Stayer X Co) Dropout X Co) X 2.-IGd (90,8) 210(9.2) 2.121 (87 d) d 10 (12 7) 2.05 5 (78 9) 550 (21.1) G ender [toy 1.170 (90 M 12d (9 5) l.Odl (85 11 180 (1-1.9) 990 (70.7) dOd(2d di < ril l LIST (91 d) 1 I d (8.7) 1.09d (89.-1) l it) (10 0) 1.057 (81 d) 2 Id 118 7) P C ellin' P J . 0 0 ” * 001** Group Control 1.152 (89.2) 110 (10 8) 1.010 (80 d) 102 (Id 7) 990 (70.0) 502 (2d I) Program 1.211 (92 t) 100 (7 0) 1.108 (88.2! 1 18 1 I 18) l.OOd (811) 2 18 (18 9) P fillin ' 00.',** I t .00',** Kthnicity White 975(919) 80 18 11 872 (80.7) l i t ( Id d) 8-11 (79 d) 220(20 7) African 01 (85 9) 10(1 11) 51 (82 d) 11 (17 7. 50(70 li 21(29 0) American Hispa me 70-1 (89 7) 81 1 10 d) 0.59 (88.1) 80(11.9) 018 (78 7) 107(21 d) Asian 595 (9-1.1) 25 (0 0) 550 (87.7) 50 (12.d) d 15 (82.1) 75 (17 9) American 18 (78,5) 5 (21.7) Id (08.-1) o (di. 0) 12 (52 2) 11 i 17 8) Indian Other 208(90 8) 21 (9.2) 190 (90.1) 21 (9 9) 187 (81 7i 12 (18 d) P viilui' .01** os*- .00.',** SMS t Low 2 10 (8d.0) 17 110 1) 199(80 9) 17 (19 1) 19d (07 di 9 1 052,8) Meddle S.dd (911) 82 (9.0) 70d (88 d) 101 (117) 7 52 (800) 18 5 (20 0) High 1.290 (92.1) 111 (7 9) 1.102 (87,8) 102 (12.2) 1.128 (80 5) 27 5 (19.5) P t nluc < .001*** .000** < 001*** S m ok in g Yes 112 (85.0) 09 ( 1 1 1) 151 (810) 102 (18 1) >27 (08.0) 15 1 052 0) Xo 1.912 (92 2) 105 (7 8) 1,595 (88.7) 20 1 (lid) 1.720 (81 0) .587 (18-1) P ( n/lit’ < .001 <001 < 001 Note, t SIvS--socioeconomic sta tu s at school level, w hich w as me a s u re d by proportion of st ude nts receiving free m eals : low=nu>re t h a n 2/d. m iddle= betw een 1/d and 2/d. high=dess th a n 1/d. + p < .10: * p < .05: ** p < .01: *** p < .001 for c h i-sq u are tests. Adolescents in the control group were more likely to drop out from time 1 to time 2. hut not from time 2 to time 3. Boys were more likely to drop out from time 2 to time 3, but not from time 1 to time 2. Compared to those who completed all surveys at three waves, the adolescents who either dropped out completely since baseline, or dropped out at one of the R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth er reproduction prohibited w ithout perm issio n . 8 8 last two follow-ups were more likely to be boys, in the control group. Afri can-Americans. American Indians, in low socio-economic status, and smokers at baseline. To further examine whether the differential attrition existed in our data, the percentages of dropouts at time 2 and 3 were calculated at each group of the prevention program and the smoking status at baseline (Ta ble 5.2). Smokers were more likely to drop out from the prevention pro gram than non-smokers in both control (15.5% vs. 9.17%. p = .003 at T2: 19.9% vs. 12.3%. p = .003 at T3) and program groups (13.0% vs. 6.50%. p = .001 at T2: 20.2% vs. 10.0%. p < .001 at T3). Table 5.2 Percentages of dropouts by prevention program and smok ing status at baseline Time 2 Time d Sm oker Non- smoker Odds Ratio: Sm oker Non- smoker O dds Ra tio Control 15.5*** 9.17 1.61* 19.9*** 12.d 1.08** Program ld.0*** 0.50 2.01** 20.2*** 10.0 2 20*** Group ■ S m ok e interaction .10 (.61) 1.19 (.58) .2-1 (,d9) i.d2 odd) (p value)' Note. t. Tests for interaction between prevention group and smoking status at base line were conducted using logistic regression analyses (dependent variables = dropout status at T2 and Td. independent variables were prevention group, smoking status at baseline, and the interaction term between these two vari ables). Odds ratios were adjusted for gender, ethnicity, and socio-economic status. * p < .05: ** p < .01: *** p < .001. R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 8 9 No interaction effects between prevention group and smoking status at baseline on the proportion of dropouts were detected (p > .10). indicating that there were no differential attritions in prevention groups in our data. 5.1.2 Prevalence rates of smoking The error rates of lifetime smoking defined as the percentages of subjects who reported "yes" for having smoked in their lifetime at a prior time point, but reported "no" at a later time point were 19.9% at wave 2 and 20.5% at wave 3. The errors of inconsistency for lifetime smoking were corrected before the calculations of prevalence rates and forthcom ing data analyses. The corrections were made based on a subject's an swer to the lifetime smoking variable at a prior wave on the original scale (1 to G). For example, if subjects selected an answer that was at a lower level at wave 2 or 3 than that at a prior wave, the answer of the later wave was replaced by the value of the prior wave to make the answer consistent across time. Overall, at the beginning of seventh grade, the prevalence rate of lifetime cigarette use was 15.97%, current cigarette use was 3.23%. and the regular/addictive smoking was 1.27%. These prevalence rates increased over time (Table 5.3). R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 9 0 Table 5 . 3 Prevalence rates of cigarette use over time (°o ) C ha ra cter Lifetime use ' C urrent use : Regular/Addictive use * 7"'(F> 7th (Sj 8th (S) 7,h (Fi 7,h (S) 8,h (S) 7lh (F) 7th (S) 8th (S) S ch ool District Anaheim 20. I f do.so' dii.so-' l.Oli-' 0 92 ' 7 Id 1.09 d 10' 2.07 Irvine 8.05' id.9d' '■ IS 2G■ 1.01' -1.19' 1.88-' 0 8 1 1,85' 2,51 Xewport- Mesa 17 IG 27,12'' dd 9d*’ 2.95 0.07 9,52' 0.90 d i d 5 00 G e n d e r Female 12.51 — 19 77'" 2 1.79'" 2,50* .5.78"' 5.78' 0 70' 1,51'" 1 70" Male 19 GG d0 95 ■57.15 5 95 8 21 8,59 1,81 l 12 1 02 Ethnicity White Id Id'1 20,SI' 27 19" 2 52 ' 5,52' 7.95' 0 SI' 2 99 .5 2d African American 20,111 ’ ■10.82' h 11.90'" 1 OS 12 21' > ’ (5 12 1 08' 0 12' 2.0 1 Latino 22 09 ' 1 10 05 12 90' -1.5S-' 7 01' 7 80'' I d I 2 91 2 12 Asian 9 9 11 '■ 1 1-1.29'" 17 78h ■ 2 02 d 21' d 80 ''' I 75 1 17-' '■ 2 92 Others 17 1-r 25 00h 27 70'' 2 8 1 7 5 1 1 7 51 1 Id 172'’ d 29 G r o u p Control 17 h i 28.G1"* d 1 d f 2,8d 0.98* 8 11* 1 (52 5 (5 1 * d 8 1" Program l l , Id 22.01 27,17 d.oo 1 99 5.95 0 9 1 2.20 1,88 Total 15.97 25.19 .'50,78 :5.2d 5.95 7.1 1 1.27 2.92 2.85 N’otc. F = fall semester: S = spring semester. t. Lifetime use was defined as having smoked part or all one cigarette in sub ject's whole life. +. Current use was defined as having smoked p art or all one cigarette in the past .10 days. §. Regular/Addictive use was defined as having sm oked more than 100 ciga rettes and smoked cigarettes in the past '50 days, or sm oked more than 10 ciga rettes per day. or used tobacco within the first .'50 m in u tes of waking. * . Significance tests were conducted by each ch a racter for cigarette use for each time point separately Values with the same letters were statistically different at a - .001 level. a-d: The proportion with sam e letters were statistically significant. + p< 10: * pc.Od; ** p<01: "** p<001. Irvine school district had lower prevalence rates of lifetime and current cigarette use than Anaheim and Newport-Mesa school districts. R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n. 91 There were significant differences in the prevalence rates of regu lar/additive smoking between Irvine and Anaheim school district only at Time 2. The differences among school districts became smaller when the subjects grew older. Male subjects had significant higher prevalence rates of all three smoking categories than females over time. Hispanic and African Americans had significantly higher prevalence rates of life time and current cigarette use than Asians and Whites. However, the differences in prevalence rates of regular/addictive smoking among dif ferent ethnicities decayed over time. The program group had signifi cantly lower prevalence rates of cigarette use than the control groups at both follow-up surveys. Xo significant differences in prevalence rates of cigarette use were detected between the control and program groups at baseline. 5.1.3 Comparison of smoking prevalence with California, na tional, and Florida samples To compare the smoking prevalence rates with that of the Califor nia. national, and Florida samples, the adolescents in the control group were selected. As shown in Table 5.4, the lifetime smoking prevalence R e p ro d u c e d with perm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n. 9 2 rates in the control group of our sample were lower than a representative California sample, in general, and this was especially true for White and Asian Americans. In addition, the current smoking prevalence rates as defined as having smoked part or all one cigarette in the past 30 days in the control group of our sample was similar to the national sample, but lower than the California and Florida samples. Table 5.4 Prevalence rates ("«) of lifetime and current (past 30 days) smoking among eighth grade adolescents with different ethnicity between California, national, and Florida sam ples Ethmcitv Lifetime smeiking 1 Current smokiing (past 30 days) i SC ' CA - SC CA National * fMorula ' (n-f)9()l in =5.870) (n=9!)0) (11=5.870) (1 1 = 15.058) in = 10.268) White 2.S. 5 16 0 S3 19 0 8 8 16 1 African A m erican 38.5 30 0 1 1 10.0 9 0 8 5 Hispanic 17 9 53.0 10 0 20.0 1 10 16 1 Asian 17 3 12 0 3.2 1 t o N A N/A T o t a l 34.3 46.2 8.4 18.1 9.2 11.9 Note. a. Lifetime use was defined ais having s moked part or all one cigarette in an ado- lescent's whole life. b. C urrent use wa.-- defined as having smoked part or all one cigarette in the past .'30 days. t. SC = Southern California. D ata were draw n from the control group (n=990) at eighth grade in our data to represent southern California in 2000. The percent ages of ethnicity in control group of our data were 39.-1% White. 2.(33% African American. 35.5% Hispanics. 12.S% Asian, and 9.7% others. +. CA = California. Data were cited from a representative sa m p le of eighth grade California youth during 199G-1997 school year (n=5.870).-IM The percent ages of ethnicity in this sample were -10% White. 9% African A m erican. 25% His panic, 17% Asian American, an d 8% multi-ethnic. §. D ata were average smoking prevalence rates in middle school (grades G - 8) from the National Youth Tobacco Survey (NATS) in 1999.-'"' D ata were average smoking prevalence rates in middle school (grades G - 8) from the Florida Youth Tobacco Survey (FYTS) in 1999.-’ R e p ro d u c e d with p erm issio n of th e copyright ow ner. F u rth e r reproduction prohibited w ithout perm issio n . 5.1.4 Change of smoking stages 9 3 In general, when the subjects started their seventh grade, there were 84.1% non-smokers. 8.6% triers. 4.5% experimenters. 1.6% current smokers, 0.4% regular smokers. 0.6% addictive smokers, and 0.3% quit ters. The proportions of nonsmoking stages decreased, whereas other ad vanced smoking stages increased over a period of two years (Figure 5.1). 8 > 100% 80% 60% c « o £ 40% 20% 0% 7th (Fall) 7th (Spring) Grade 8th (Spring) ■ Addicted Sm oker & Regular Sm oker 0 Current Sm oker □ Experimenter 3 Quitter □ Trier □ Non-smoker Figure 5.1 Change of sm oking stages. P e rc e n ta g e s of stages from n o n -sm o k e rs to ad d icted sm okers were: 84.1% , 8.6%. 0.3%. 4.4%. 1.6%. 0.4%. a n d 0.6% at 7th g r a d e (fall), 74.8%. 1 1.6%. 1.0%, 7.3%, 2.8%. 1.1%, a n d 1.4% at 7th g rad e (spring), a n d 69.4%. 12.8%. 2.4%. 9.3%, 3.5%. 1.3%. a n d 1.3% at 8tU grade (spring) respectively. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 5.1.5 Patterns of regular/addictive smoking 9 4 Overall 1.3% of adolescents reported that they smoked more than 100 cigarettes in their whole life. 0.73% reported that they smoked within 30 minutes of waking, 0.34% smoked more than 10 cigarettes per day at the start of seventh grade. The prevalence rate of regular smoking de fined as having smoked more than 100 cigarettes, or used tobacco within 30 minute of waking, or having smoked more than 10 cigarettes per day was 1.01%. Prevalence Rates of Addicted Smoking ~ S m ok ed 100 cigarettes H S m ok ed within 30 mm of waking B S m ok ed 10 cigarettes per day tS A ddicted sm okers ■ FTND s c o r e >= 1 7th (Fall) 7th (Spring) 8th (Spnng) Grade Figure 5.2 In c re asin g trend of regular/addictive sm oking Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. The prevalence rate of addictive smoking defined as Fagerstrom Test for Nicotine Dependence (FTND) score greater than 1 was 0.88% (Figure 5.2). A significant increasing trend of prevalence rates for regu lar smoking (Cochran-Armitage Z score = - 6.30. p < .001) and addictive smoking (Cochran-Armitage Z score = -1.73, p = .04) was observed. 5.1.6 Intraclass correlation (ICC) and design effects The intraclass class coefficients (ICCs) and design effects were cal culated to measure the degree of clustering for smoking related variables in this study. As shown in Table 5.5. in general, the intraclass correla tion coefficients of lifetime smoking (ICC ranged .023 ~ .034) were larger than monthly smoking (ICC ranged .008 - .011) and daily smoking (ICC ranged .000 ~ .000) on a continuous scale. Larger intraclass coefficients were observed when lifetime and monthly smoking variables were di chotomized. With an average cluster size of 115.07 in this study, the de sign effects ranged from 1.08 to 4.88 for continuous variables. For di- chotomous regular and addictive smoking variables, the ICCs were small at time 1. large at time 2 (ranged from .034 to .127). and sporadic at time 3 (ranged from 0 to .135) at time 3. The design effects ranged from 1 to 15.49 given an average cluster size of 115.07 in this study. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 9 6 Table 5.5 In traclass correlation coefficients (ICCs) and design ef fects for sm oking related v ariab les Measured outcome Variable ICC’+ Design effect* T1 T2 T2 T1 T2 T2 Continuous Lifetime smoking 0.022 0.021 0.021 2.(52 -1.88 1.5 1 Monthly smoking 0.00s 0.011 0.006 1.91 2.26 1.68 Dailv smoking 0.00(5 0.009 0.007 1.68 2.02 1.80 Dichotomous Lifetime smoking 0.075 0.0S1 0.07 9.55 10.21 8.98 Monthly smoking 0.107 0.051 0.02 1 12.21 6.82 .171 Smoked > 100 ciga rettes in lifetime 0.00 0.127 0 125 1.00 15.-19 16,1 Smoked C 10 ciga rettes per dav 0.00 0.02 I 0.02 1.00 1.88 1.12 L’sed tobacco within 20 minutes of waking 0.00 0.115 0.00 1.00 1 1.12 1.00 Regular smoking 0.00 0.011 0.022 1.00 5.68 1.76 Addictive smoking (FTND score > 1) 0.00 0.08(5 0.00 1.00 10.81 1.00 Note. X = 2. 052. t. ICCs w ere e s tim a te d by the following form ulas: r , < r P -------------------------------- — - n , = — : — ------- . con tin u ou s r - + q- binary & ~ /T 3 +. Design effects w ere e s tim a te d by th e following formula: design effect = 1 + ( 777 - 1 ) • p ut : in (average cluster size)= l 15.07. as calculated by HI = 1 m , ' / l ni.. 5.2 Latent Transition Analyses of Sm oking Stages There were three sets of parameters that were estimated in latent transition models: measurement parameters, unconditional probabilities of latent status membership, and transition probabilities. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 9 7 5.2.1 Cross-validation Before the analyses were conducted, the total sample was ran domly divided into Sample A and Sample B. Separate analyses were per formed on each sample and the stability of the findings were compared. Nearly perfect stability of our sample was observed. 5.2.2 Measurement param eters (pparameters) The measurement parameter estimates of latent status were con strained to he equal across time and group. As shown in Table 5.(5, the measurement parameter estimates suggested overall adequate reliabil ities of measured variables in our data. For the first latent status, i.e. non-smokers, all of the parameter estimates were equal to zeros, indicat ing that the probabilities of answering "yes" to all four measured vari ables by the subjects were consistently low. For the last latent status, i.e.. addictive smokers, all of the parameter estimates were equal to one indicating the probabilities of answering "yes" to all four measured vari ables by the subjects were consistently high. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 9 8 T able 5.6 M easurem ent (p) p a ra m ete r estim ates of la te n t status in relation to m e a su re d variables * Measured variable Latent status Lifetime smoking Monthly smoking Daily smoking Addictive smoking Non-smokers 0.0 0.0 0.0 0.0 Tners/Expenmenters 1.0 0.0 0.0 0.0 Current smokers 1.0 1.0 0.0 0.0 Regular smokers 1.0 1.0 1.0 0.0 Addictive smokers 1.0 1.0 1.0 1.0 Note. t. The m e a s u re m e n t (p) p a r a m e t e r estimate.-? of la ten t s ta tin < in relation to m easured v ariables w ere constrained to be eq u al acro ss three tim e points. X=2.05:i. For other latent statuses, the parameter estimates were also ade quately reliable such that the probabilities of answering "yes” to the measured variables that most likely reflected the latent status were equal to one and the other measure variables that least likely reflected the latent status were equal to zero. Good model Fit was also obtained as suggested by the ratios of G- statistic to the degree of freedom the entire sample and the sub-groups. 5.2.3 Unconditional probabilities of latent class membership Overall 88.5% of subjects were in non-smoker latent status, 9.3% in trier/experimenters latent status, 1.4% in current smokers latent status, 0.5% in regular smokers latent status, and 0.3% in addictive Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 9 9 smokers latent status at Time 1. The probabilities in the non-smoker la tent status decreased, while the probabilities in other latent status in creased over time (Table 5.7). T able 5.7 U nconditional prob ab ilities (5 param eter) of laten t statu s an d predicted n u m b e r of subjects for e n tire sam ple a n d by group L a te n t s ta tu s P re d icted probability P redicted n u m b e r of subjects Tim e 1 T im e 2 T im e 3 Tim e 1 Tim e 2 T im e 3 T o ta l (N =2.053) N o n -sm o k ers .885 .821 .755 1817 1092 1550 T rie rs /E x p e rim e n te rs .098 .129 .172 191 205 353 C u r r e n t sm o k e rs .011 .027 .012 29 55 80 R e g u la r sm o k e rs .005 .010 .017 10 21 35 A ddictive sm o k e rs .003 .009 .011 0 18 29 C ontrul (n=990) N o n -sm o k ers .87 1 * 1 .797**1 ' .729** 802 789 722 T r l e r s/ H x p e r l m e n t e r s . 109**1 .1 15**- .185 108 111 183 C u r r e n t sm o k e rs .008**'' .029 .011 s 29 11 R e g u la r sm o k e rs .007 .015 .021**- 1 15 21 A ddictive sm o k e rs .005- .01 l**h .018- 5 13 19 P ro g ra m <n=l.0G3) N o n -sm o k ers .898*' .S-19**1 ’ .779**1 ' 955 902 828 T r le rs/ E x p e r i m e n t e rs .078**'1 .111**- .100 83 121 170 C u r r e n t sm o k e rs .019**' .020 .0-11 20 28 -13 R e g u la r sm o k e rs .00-1 .007 .012**“ - ' -I i 13 A ddictive sm o k e rs .ooi- ,00-l**h .008* 1 -1 9 N'ote. P robabilities w ith identical le tte rs w ere significantly different across th e control an d p ro g ra m groups using 2 v 2 contingency table y--tests. -T P<10: * p< 05: ** p<0l. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 0 In the program group, there were higher percentages of nonsmok ers and current smokers, but lower percentages of trier/experimenters, regular smokers, and addictive smokers than in the control group. 5.2.4 Transition probabilities Overall transitions across latent statuses were stable for non- smokers. trier/experimenters, and addictive smokers (see Table 5.8). For example, 93.2% of non-smokers at time 1 remained non-smokers at time 2. and 91.0% of non-smokers at time 2 remained non-smokers at time 3. Few non-smokers at time 1 progressed to trier/experimenters (4.9%), current smokers (1.4%). regular smokers (0.3%). and addictive smokers (0.2%) at time 2. Although 82.0% of trier/experimenters at time 1 remained trier/experimenters at time 2, the transition probabilities to current smokers (9.9%), regular smokers (3.0%). and addictive smokers (3.9%) were higher than non-smokers. Among current smokers at time 1 or 2. more than half regressed to trier/experimenters at Time 2 or 3. but reasonable numbers of them remained current smokers at time 2 (17.7%) and time 3 (33.4%). Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 1 Table 5.8 T ran sitio n probabilities (x p aram eter) betw een latent statu ses for en tire sam ple a n d by group L a t e n t s t a t u s a t T r a n s i t i o n fro m T i m e 1 to T i m e 2 L a t e n t s t a t u s a t T i m e 2 T r a n s i t i o n fr o m T i m e 2 to T i m e -1 L a t e n t s t a t u s a t T i m e ■ ! tittle 1 N -S T -K -S C -S R-S A -S N -S T -K -S C -S R-S A -S N o n - s m o k e r s (N -S) 0 1 2 0 1!) O i l d o t a l iN = 2.07).li 00.1 0 0 2 f) HI 07,7 0 1 7 007, 007, T r ie r / ( e x p e r i m e n t e r s (T-K -S) o szt; of)!) 0 1 0 Of!) 0 S I 7 127, 017, 0 2 1 C u r r e n t s m o k e r s (C -S i 0 '.71 177 177 071 0 7)10 1.11 I t 02 R e g u l a r s m o k e r s |R -S i 0 271 .11,1 1S2 1S2 0 2 2 0 2 0 2 17,1 2 IS A d d ic tiv e s m o k e r s (A -Sl 0 0 107 107 0 0 7 0 .172 10S 10.1 17,7 N o n - s m o k e r s i.N-Si ! > 1 7 ," ' 07.7 0 I f ) - 1 - C o n t r o l (n =!)!)()) 007.*- 00.1 01 1 07,0 O IS 007, 0 0 0 T r i e r K xpe r im e ii te i 's (T -K -S , I) s [ s O S!)*1 0 1 7 0 1 0 * " 0 701 *** 12 1 O', 1 * 0 I ','* * C u r r e n t s m o k e r s tC -S i I) !. ,* " ■ 1 2 V * *'• 17',*** 127,*** 0 (.22*** : 217**** 100* 07,2 R e g u l a r s m o k e r s ( R-Si 0 12!) 2 S 0 * " ' III***' 1 1 t ....... 0 2 0 S * * * ’ 2 0 0 * * * " IS 10 !*♦** A d d ic tiv e s m o k e r s lA -S i 1 ) 1 ) 0 2 S 0 .121 ***' 101 IU 1 * .17,7***' N o n -s im ik e r s tN -S i ;) to* *1 o n o n — ‘r u g r a i n tn = 1 00 :ii 00 1*' 00 1 0 1 7 07,!) 0 1 0 007, 00.1 T r ie r ; K xpe r i m e n t e i s (T -K -S i 0 S 2 7 1 2 .!* 1 0 1!) 0 1 2 * * * 0 S I S " * " 127 01 7** * ■ 0 0 S * * ’ ; C u r r e n t s m o k e r s (C -S) o o.",i***s goa***'i ()!)«***• 0I!)***> 0 :IH!)*** i 1.12* *r 1 11**- 0.10 Regular smokers (R-S) 0 0 7.1*** * 2 l*.***! 2 17,***'" 0 .107,***' Li!)*** ‘ 1.1!)*** • 1 1 7 * * '* A d d ic tiv e s m o k e r s (A -S) 0 0 I 0 0 0 7,1.1***' 0 27,***- 27,***' Note, a-z: the comparisons of transition probabilities were made between the control and program groups with the same letters indicating significant differences us ing 2<2 contingency table y--tests. + p<.10: * p<.05: ** p<.01: *** pc.OOl. Relatively large number of current smokers progressed to regular smokers (17.7% at time 2: 13% at time 3). or addictive smokers (7.1% at Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 2 time 2: 2.0% at time 3). Nearly 20% of regular smokers progressed to ad dictive smokers both from time 1 to time 2 and from time 2 to time 3. 18.2% (from time 1 to 2) and 3 5.3% (from time 2 to time 3) remained regular smokers. Nearly half of regular smokers regressed to either cur rent smokers or trier/experimenters. Addictive smokers seemed to be more persistent to remain at the same level over time (06.7% at time 2: 35.7% at time 3). although many of them reduced their levels of smoking. 5.2.5 Omnibus tests of transitional probabilities The omnibus tests for overall patterns of unconditional probabili ties (rf parameters) and transition probabilities (r parameters) between two groups (control vs. program, female vs. male, and White vs. non- White) were conducted in a fashion of nested models. A basic model (Mi) with free parameter estimates for unconditional probabilities or transi tion probabilities between two groups was established and the model fit index (G: with given degree of freedom) was obtained. Three alternative models with parameter estimated constrained to be equal for uncondi tional probabilities (M2). transition probabilities from time 1 to time 2 (Mi 2) and transition probabilities from time 2 to time 3 (M2. 1) were estab lished and the model fit indices were calculated (see Table 5.9). Because Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 3 the basic mode and the alternative models were nested in nature, the dif ferences of the G~ statistics and the degree of freedom between the basic and alternative models were used to examine whether the overall pat terns of unconditional and transition probabilities between two groups were statistically significant hy using tests. The results showed that overall patterns of proportions of latent status were significantly different between control and program groups (p = .006). The general patterns of transition probabilities from time 1 to time 2 were marginally significant between control and program group (p = .006) and significant between White and non-White (p = .003). The overall patterns of transition prob abilities from time 2 to time 3 were statistically significant between White and non-White (p = .001). Although the omnibus tests showed insignificant differences for the overall patterns of transition probabilities from time 2 to time 3 be tween the control and program groups, the overall patterns of uncondi tional and transitional probabilities between males and females, and the overall pattern of unconditional probabilities between Whites and non- Whites. some specific probabilities might be statistically significant be cause the omnibus tests in LTA might not be particularly sensitive for accounting for these specific differences (personal communications with Dr. Linda Collins and her LTA software programming staff). Thus fur Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 4 ther analyses for testing differences in specific probabilities between two groups were carried out using a 2 x 2 contingency table approach. Table 5.9 O m n ib u s test for unconditional p ro b ab ilities (5) an d tr a n sitio n a l probabilities (t) betw een g ro u p s Model' C r (D. f. •) \C7: (.\n. f.) P value Prevention group (program versus control) Mi 2 19.27 (811(1) - M j 2(1(1.72 (8120) i i.i5 in .00(1*** Mu 272.99 (81(12) 2(1.72 (1(1) .090+ Mj, 20 t.(1 (1 (81(12) 15.07(10) .520 (lender (male versus female) M: 2(10.0.1 (811(1) - M j 2-1(1.97 (8120) 1.(15 (1) .(1(11 Mu 2(12.88 (81(12) 2(1.2(1 (1(1) .107 M j i 2(10.9:') (81(12) 2 1 .(K 1 (1(1) . 10(1 Kthnicity (white versus nonwhite) Mi 227.59 (811(1) - M j 2(1(1.(17 (8120) 5.77 (1) .217 Mu 2(1(1 78 (81(12) (10.10 (10) .o o : r * M j i 270.75 (81(12) -id. id d o ) .000(1*** N ote. t . Mi = B asic m odel w ith no constraints (2 la ten t c la sse s and a latent sta tu s). M j = M odel w ith c o n s tra in ts on unconditional p ro b a b ility e s tim a te s across tw o groups. M u = M odel w ith c o n s tra in ts on tran sitio n p ro b ab ility e s tim a te s b etw e en tim e 1 a n d tim e 2 across two groups. M ji = M odel w ith c o n s tra in ts on tran sitio n p ro b ab ility estim a tes b etw e en tim e 2 a n d tim e :f across two groups, i. I). F. = D eg ree of freedom . + p < .10: ** p < .01; *** p < .001. 5.2.5.1 Comparison of transition probabilities by prevention group Because the current version of LTA program does not enable com parisons of specific probabilities between two groups, a scries of 2 x 2 con tingency tables were generated and tested using — tests. In general, as Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 5 also shown in Table 5.8. the program group had higher probabilities of remaining in the same latent statuses and lower probabilities of progress ing to more advanced latent status. For example, compared to the control group (91.6%). the program group had higher a probability (9 1.7%) of re maining non-smokers from time 1 to time 2. Less trier/experimenters in the program group progressed to addictive smokers than in the control group at time 2 (1.3% vs. 6.2%) and time 3 (0.8% vs. 3.6%). Less current smokers progressed to regular smokers (9.8% vs. 37.5%) or addictive smokers (4.9% vs. 12.5%) at time 2. whereas more current smokers re gressed to trier/experimenters at time 2 (65.6% vs. 42.9%). In contrast, more regular smokers progressed to addictive smokers in the program group than the control group both at time 2 (25.0% vs. 14.3%) and time 3 (41.8% vs. 8.8%). More addictive smokers regressed to current smokers at time 2 (100% vs. 0%) and trier/experimenters (51.3% vs. 33.0%) or regular smokers (24.4% vs. 13.7%) at time 3 than the control group. 5.2.5.2 Comparison of transition probabilities by gender Overall females had higher probabilities of remaining at the same latent statuses or regressing to lower level of latent statuses and lower probabilities of progressing to more advanced level of latent statuses as shown in Table 5.10. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 6 Table 5.10 T ransition probabilities betw een laten t statu ses by g e n d er L a t e n t .s ta t u s a t T r a n s i t i o n fr o m t i m e 1 to t i m e 2 L a t e n t s t a t u s a t t i m e 2 T r a n s i t i o n f ro m t u n e 2 to t u n e 3 L a t e n t s t a t u s a t tim e 1 t i m e I N -S T -K -S C -S R-S A -S N -S T - K -S C -S R-S A -S N o n - s m o k e r s (N'-S) 9-i n* ... 0 4 1 -i, 0 0 8 * * ' F e m a l e s 0 0 2 I n = 1 .0 5 7 1 0 0 2 9 2 9 * 1 C i. 1 .1 0 1 4 0 0 1 **'*■ 0 0 2 * " T r i e r ' K x p e r i m e n t e r - (T -K -S i 0 8 2 8 1 0 5 0 5 5 * * " 1 012**** 0 8 0 1 14 1 *v 0 4 1 0 1 6 * ■ ( C u r r e n t s m o k e r s (C -S ) 0 7 7 4 0 7 5 * * * s 151 — ■ 0 0 5 2 5 3 3 3 143 0 R e g u l a r - m u k e r s (R -S) 0 403***- 3 9 8 * * ' 1 9 9 -k 0 0 2 7 3 * * ' 1 8 2 3 6 3 1 8 2 - ' A d d ic t iv e s m o k e r s (A -Si 0 0 3 3 3 3 3 3 3 3 3 0 0 0 5 * *"’ 5 * * *-■ N o n - s m o k e r s (N -S ) 9 1 6 * * •■ 0 5 7 —' 0 2 1 — M a le s 0 0 3 H l=‘)9(il 0 0 2 9*: 0 6 2 0 2 0 0 9** '" 0 0 8 * " T r i e r / K x p e r i m e n t e r s (T-K -S ) 0 8 1 6 1 0 2 0 .35 ***'1 047**** 0 8 3 - ' 1 12*)’ 0 3 0 0 2 8 * 1 C u r r e n t s m o k e r s (C -S ) 0 3 9 7 ' " ' 2 7 2 * * * s 198**'- 1 3 2 0 4 9 4 3 2 0 1 17 0 6 9 R e g u l a r s m o k e r s (IC S) 0 165***- 34**.' 1 6 5 * k 3 3 0 2 0 0 * * ' 1 8 7 3 7 5 2 3 8 - ' A d d ic tiv e s m o k e r s (A -S) 0 0 0 0 1 0 0 0 4 7 6 1 5 9 07 9 * * * ' 2 8 6 * * * " Note, a-u: the comparisons of transition probabilities were made between females and males with the same letters indicating significant differences using 2-2 contin gency table /--tests. + p< . 10: * p< 05: ** p < 01: *** p< 001. For example. 94.6% of female non-smokers at time 1 remained non-smokers at time 2, whereas only 91.6% of male non-smokers re mained at the same level at time 2. Female trier/experimenters were less likely than males to progress to addictive smokers (1.2% vs. 5.6% at time 2). however, were more likely to progress to regular smokers (5.5% vs. 3.12% at time 2). In contrast, more female current smokers reduced their Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 7 smoking levels to non-current smoker (77.4%) than males (39.7%): more female regular smokers regressed to trier/ experimenters or current smokers than their male counterparts. However, female additive smok ers were more likely to remain at the same latent status at time 3 than males. 5.2.5.3 Comparison of transition probabilities by ethnicity In general. White adolescents were more likely to remain at the same latent status, hut less likely to reduce their smoking levels to lower latent status, and less likely to progress to more advanced latent status as shown in Table 5.11. For example. White non-smokers had smaller transition probabilities to progress to current smokers from time 1 to time 2 than non-Whites (0.8% vs. 1.9%). White trier/experimenters had smaller probabilities of progressing to addictive smokers (1.5% vs. 4.9%). but had larger probabilities of progressing to current smokers (12.7% vs.8.G%) or regular smokers (7.5% vs. 2.2%). White current smokers had higher probabilities to remain at the same status (25.8% vs. 12.4%) and progress to regular smokers (24.5% vs. 12.4%). but had lower probabili ties to regress toward trier/experimenters (41.5% vs. 68.9%). The same was true at time 3. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 8 Table 5.11 T ran sitio n p ro b ab ilities am o n g laten t statuses by eth n icity L a t e n t s t a t u s a t T r a n s i t i o n fro m t u n e 1 to tim e L a t e n t s t a t u s a t t i m e 1 , •> T r a n s i t i o n fro m t i m e 2 to ti m e 3 L a te n t s t a t u s a t t i m e 3 t i m e 1 N -S T -K -S C -S R -S A -S N -S T -K -S C -S R-S A -S N o n - s m o k e r s (N -S ) 9-40*1 0 4 0 0 8 - ' W h i t e 0 0 3 (n = S : ta | 0 0 3 9 1 4 0 4 9 0 2 2 0 0 7 0 0 7 T r i e r K x p e n m e n t e r s iT - K - S i 0 7 8 4 "* * ' 1 0 7 5 — 0 1 5 * * *r 0 8 0 6 1 15 0 4 5 0.34*" C u r r e n t s m o k e r s 1C - S 1 0 4 15***i 2 5 8 * * * 1 ’ 2 4 5 * * * ' 0 8 2 - 0 5 0 6 2 2 3 — f 16 7 **‘i 1 0 5 R e g u l a r s m o k e r s IR - S i 0 2 5 -s 2 5 9 * * *tl: 4 9 1 0 0 3 3 5 * * " .3.32 166***' 166***'' A d d ic t iv e s m o k e r s (A -S i 0 .585 3 9 9 2 1 6 0 0 0 o***-. 5 **"•'• 5* .., N o n - s m o k e r s l.\'-S> 9 2 I * ■ • 0 5 5 0 1 9 — ' N o n - W h i r 0 0 3 e 1 n = 1 2 1 Si 0 0 2 9 1 7 0 6 3 01.3 00.3 0 0 .3 T r i e r K x p e r i m e n t o r s (T -K -S ) 0 842 *** ' 0 8 9 *M 0 2 9 ....... 0 4 1 ***' 0 8 2 2 1.31 0 .30 0 1 7 - C u r r e n t s m o k e r s (C -S i 0 6 8 9 1 2 4 * * * h 124***' 0 6 2 +i 0 .5 1 4 378 *** !’ 108**'i 0 R e g u l a r s m o k e r s I IT S ) 0 2 8 6 -k 4 2 9 * * * n- 0 2 8 6 0 106***| 0 5 9 6 * * * ' 2 9 8 * * ' “ A d d ic t iv e s m o k e r s 1A - S 1 0 0 3.33 3 3 3 3 3 3 0 5 4 5 0 9 1 — 0 9 1 * —+ 2 7.3 *** ' Note, a-z: the comparison.' of transition probabilities were made between Whites and non-Whites with the same letters indicating significant differences usintt 2-2 contingency table / J-tests. + p<. 10: * P<.0r,: ** pc.01: *** p<.001. White regular smokers were more likely to remain at the same status ancl less likely to declined their smoking status to lower level at time 2: and the reverse was true at time 3. More than half of the White addictive smokers remained at the same status over time, whereas about one third of non-White addictive smokers remained the same status. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 0 9 5.2.6 Comparison of smoking stages by model and definition To directly compare the proportions of smoking stages based on LTA models and based on theoretical definitions, three stages that were defined based on theory (trier, experimenter, and quitter) were combined to form a single stage. In general, the actual proportion was lower for non-smokers than predicted proportion at each time point, and it was higher than predicted proportion for trier/experimenters. The actual and predicted proportions were similar for current, regular, and addictive smokers (Table 0.12). Table 5.12 C om parisons betw een predicted a n d actu al proportions of sm oking stages for e n tire sample (n=2,053) S m oking stage Time 1 Tim e 2 Time 3 P ("..p A Vo) P Co) A (%) P<%) A (•’,.) N on-sm okers 88.5 84.1 82,1 74.8 75.5 69.4 T riers/E x p e rim e n te rs 9.3 13.3 12.9 19.9 17.2 24.6 C u rre n t sm okers 1,1 1.6 2.7 2.8 1.2 3.5 R eg u lar sm okers 0.5 0.6 1.0 1.1 1.7 1.3 A ddictive sm okers 0.3 0.3 0.9 1.4 1,1 1.3 C hi-square test r 17.9 (-1 ykie 10.7 (-11 ie * * 35.1 ( l)*x* N'ote. t. C h i-square te st w as conducted betw een p red ic ted a n d a c tu a l m em bers ot’ stages a t th re e tim e points se p a ra te ly $. P = predicted proportion, A = actual proportion. ** p < .01; *** p < .001. The chi-square tests yielded significant differences between the proportions of actual and predicted smoking stages. Since the model- Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 1 0 based smoking stages were based on the four measured variables ad justed for measure errors, the difference in proportions between model- based and definition-based smoking stages may be due to the measure ment errors. 5.3 Growth Mixture Modeling of Smoking Trajectories To empirically examine the possible patterns of smoking progres sion, latent growth mixture modeling was conducted. First, uncondi tional growth mixture models (i.e.. models without covariates) were con structed to identify possible classes or patterns of smoking trajectories and the proportions of class members. Second, predictors were added to the model to examine the differences of class members across different levels of these predictors. Finally, the results obtained from SAS Proc Traj procedure were validated by comparing the results by Mplus pro gram. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 5.3.1 Pattern of smoking trajectories 1 1 1 The number and the order of classes were empirically determined based on the Bayesian Information Index (BIC) and Akaike Information Index (AIC). Comparing; to a two-class linear latent growth trajectory model (BIC - - 3850.43. AIC = - 3833.55). and a two-class quadratic la tent growth trajectory model (BIC = - 3828.90. AIC= - 3806.40). a three- class quadratic latent growth mixture model (BIC = - 3735.58, AIC = - 3701.82) was selected because of its smallest absolute value of BIC and AIC. In general, three latent classes of trajectories represented stable non-smokers (69.2%), slow escalators (trier/experimenters) (28.4%), and early stable smokers (2.4%) as shown in Figure 5.3. For slow escalators, they progressed from triers/experimenters towards current smokers in a slow rate. For early stable smokers, they started as current smokers, progressed to regular smokers rapidly, and decreased their use of ciga rettes thereafter. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 1 2 Latent Growth Trajectories of Smoking Censored Normal Model, Total Sample (n=2,053) S ta g e 7th G ra d e , Fall 7th G rade, S pring 8th G rade, S pring C la ss %: 1—t-t- 69.2 28.4 9 -9 -9 2.3 F ig u re 5.3 Sm oking T rajectories for the E n tire Sample. Solid lines re p re se n t observed trajectories an d d a sh ed lines rep re se n t expected trajectories. The valu es of sm oking stag e are: l= n o n -sm o k o r, 2= non-current sm o k er. 3= m onthly sm oker. 4=daily sm oker, an d 5=addictive sm o k er. T h e s a m e m ean in g s a p ply to the sim ila r figures thereafter. 5.3.1.1 Prevention groups There were more stable non-smokers (72.7%). less slow escalators (25.4%), and less early stable smokers (1.9%) in the program group than in the control groups (65.6%, 31.6%, and 2.8% respectively) (Figure 5.4). Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 113 Latent Growth Trajectories of Smoking Censored Norm al M odel, Program G roup (n=1,063) S tag e 7th G rade. Fall 7th G ra d e , S p rin g 8 th G rade. S p rin g C la s s %: * - t —f 72.7 S -9 -? 25.4 *7-9-9 1 9 Latent Growth Trajectories of Smoking C en s o re d N orm al M odel, C ontrol G ro u p (n=990) S tag e 7th G rad e, S p ring 7th G rade, Fall 8 th G rade, S p rin g C la s s %: < i t 65.6 9 31.6 9 -9 - 9 2.8 Figure 5.4 Sm oking T rajectories by P rev en tio n G roups. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 114 Slow escalators showed a gradual progression trend in both control and program groups. Early stable smokers in the control group stared at a lower level of smoking but a faster progression toward daily smoking at Time 2 than those in the program group. However, early stable smokers in both groups showed a decreasing trend after Time 2. 5.3.1.2 Gender Males had a smaller proportion of stable non-smokers ((53.1%), but larger proportions of slow escalators (34.3%) and early stable smokers (2.6%) than females (75.1%. 23.2%. and 1.7% respectively) (Figure 5.5). Similar growth rates of slow escalators were observed for both groups. However, male stable smokers progressed more rapidly toward addictive smoking at Time 2 and then declined at Time 3. while females showed a steady progression over time. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 1 5 Latent Growth Trajectories of Smoking C ensored Normal M odel. M ales (n=996) S tage 7th G rade. Fall 7th G rad e. Spring 8th G rad e. Spring C la s s %: i * t 63.1 9 9 9 34.3 9 -9 -9 2.6 Latent Growth Trajectories of Smoking C e n s o re d N orm al M odel. F e m a le s (n=1057) S tage 7th G rad e, Spring 8th G rad e. S p ring 7th G rade. Fall C la s s " /.: 1 1 1 75.1 9 9 9 23.2 9 -9 -9 1.7 Figure 5.5 Sm oking Trajectories by G ender. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 116 5.3.1.3 Ethnicity In general, the patterns and proportions of smoking trajectories varied across different ethnic groups (Figure 5.6). Asian adolescents had highest proportion of stable non-smokers (82.4%). followed by Whites (72.8%), other ethnic adolescents (72.2%). In contrast. African American adolescents had lowest proportion of stable non-smokers (55.0%). followed by Hispanic adolescents (57.1%). Hispanic adolescents had highest proportion of slow escalators (41.1%). followed by African Americans (40.9%). Asian adolescents had lowest proportion of slow escalators (15.2%). followed by Whites (24.6%), and other ethnic adolescents (26.4%). African American adolescents had highest proportion of stable smokers (4.1%). followed by Whites (2.6%). and Asians (2.4%). Relatively low proportion of stable smokers was observed in other ethnic (1.4%) and Hispanic (1.8%) adolescents. For stables smokers, ado lescents with different ethnicity showed diverse patterns of trajectory. African American and other ethnic adolescents started at an addictive level at Time 1 and declined to a level of daily smoking at Time 2 and 3. White adolescents started at a level of current smoking, progressed to a daily using level at Time 2. and stayed at that level at Time 3. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 1 7 Latent G row th Trajectories of Sm oking C e n s o re d Norm al Model. Asian (n=343) S t a g e ____________________ ______________________ 5 Latent Growth Trajectories of Smoking C en so red Norm al M odel. African A m erican (n=46) S tage 5 7th. Spring 7th. Spring 8th. Spring Latent Growth Trajectories of Smoking C en so re d N orm al M odel. W hite (n=835) S ta g e ________________ -_- __________________ 5 • 1 * - 7th. Fdil C lass %: 7th. Spring — 72.8 — 8th. Spring • 2.6 L atent Growth Trajectories of Smoking C en s o re d N orm al Model. H ispanic (n=613) S tag e 5 4 3 2 1 XI 7th, Fall 7th. Spring C la ss %: 57.1 “ r r 41.1 Latent Growth Trajectories of Smoking C en so red N orm al Model. O ther E thnicity (n=213) S ta g e ______________________________________________ 8th . S p ring C lass %: 5 4 3 2 1 \ \--- 7th. Spring 8th. Spring * 72.2 Figure 5.6 S m oking T rajectories by E thnicity. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 1 8 Hispanic adolescents started at an experiment smoking level, pro gressed rapidly toward addictive level at Time 2. and decreased rapidly at Time 3. Asian adolescents started at daily smoking level, declined to current smoking level at Time 2. but slightly increased at Time 3. 5.3.2 Predictors for differentiating smoking trajectories Odds ratios and the 95% confidence intervals of stable smokers and slow escalators compared to stable non-smokers in relation to demo graphic. intrapersonal, and interpersonal predictors were calculated si multaneously in a conditional latent growth mixture model (i.e.. a model with covariates included) as shown in Table 5.13. 5.3.2.1 Demographic predictors Male adolescents were more likely to be classified as slow escala tors (OR = l.GG. p < .001). and stable smokers (OR = 2.4. p < .10) than their female counterparts. Compared to White adolescents. African Americans were more likely to be classified as slow escalators (OR - 2.25, p < .05): in contract, Asians were less likely to he classified as slow esca lators (OR ,G7. p < .05). Adolescents in the prevention program was as sociated with lower probability of being classified as slow escalators (OR Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 1 9 = .81. p < .10). and especially lower probability of being classified as sta ble smokers (OR = .17. p < .001). Adolescents with higher socioeconomic status as defined by percentages of students not receiving free meal at school were less likely to he classified as slow escalators (OR = .99. p < .01). Ethnicity and socioeconomic status were not related with being clas sified as stable smokers. 5.3.2.2 Intrapersonal predictors Adolescents with lower academic grades had higher risk of escalat ing in smoking in that those with mostly C ’s had two folds likelihood to be classified as slow escalators (OR = 2.13. p < .001). and more than 10 folds likelihood to be classified as stable smokers (OR = 14.42, p < .001) than those who reported mostly As. Adolescents who showed intentions to smoke at baseline were at higher risk to be classified as either slow esca lators (OR = 4.92. p < .001). or stable smokers (OR = 329.44. p < .001) than those who did not show intentions to smoke at baseline. Refusal self-efficacy was only marginally related to lower risk of being stable smokers (OR = .38, p < .10). Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 0 T ab le 5.13 Odds ratios o f latent class membership in relation to predictors Slow esca lato rs vs . stable S tab le sm o k e rs vs. ;stable non- P redictor non-■smoker:s sm o k e rs O R ' (95" ii Cl ■ ) O R <9!>% Cl) Demographic G e n d e r F em ale 1.00 ” 1.00 M ale LOG” * (1.30. 2.12i 2. i + (0.86 . 6.69) E th n ic itv W hite 1.00 1.00 A frican A m erican 2 25* (1.08. 1.(591 0.68 (0.0-1., 11.15) H ispanic 1.08 (0.78. 1.51) 0. 1(5 (0.12. 1.76) A sian 0.G7* (0.-15. 0.99) 1.06 (0.21. 5,18) O th e r 0.85 (0.55. 1.-32) 0.95 (0.21. -1.28) G roup C ontrol 1.00 1.00 P ro g ram 0.81 + (0.0 1 . 1.0-3) 0.17** (0.05. 0.55) S E S * Low 1.00 1.00 H igh 0.99** (0.99. 1.00) LOO (0.97. 1.02) Intrapersonal G PA * M ostly As 1.00 1.00 M ostly Bs 1.05*** 11.20. 2.17) 1.27 (0.35. -1.56) M ostly Cs > 13**’ d .3 9 . 3.28) 1 1,12*** (-3.79. 5-1.88) In te n tio n to sm oke No 1.00 1.00 Yes J C J-> ’r** (3.-15. 7.02) -329,1 1*** (3-1.31. 3163.0) Re fusa 1 se 1 f-e fficac v No 1.00 1.00 Yes 0.91 10.0)9. 1.19) 0,38+ (O.i-3. 1.12) Interpersonal P a r e n ts sm oking N one 1.00 1.00 1 p a re n t 1.58** (1.20. 2.09) -1.8-4* (1.36. 17.23) 2 p a r e n ts 2.56*** (1.78. 3.69) 1,11* (1.11. 17.3-1) F rien d s' sm o k in g * N one 1 00 1.00 1 friend 3 19*** (2.11. -1.83) -1.09+ (0.88. 25.0-1) 2-3 friends 0.0 i (2.2-3. 5.7!) 28 13*** (0.30. 12-1,52) -1-5 friends 3.63* (1.2-1. 10.6-1) 101,13*** (21.28. 1221.6) N ote: t. O R - odds ratio, f t . R eferent group, i. 95% Cl =95% confidence in te rv a l of odds ratios. §. SES = socio-econom ic sta tu s, which w a s defined by (100 - p e r c e n ta g es of stu d e n ts h a v in g free m eal at school)/10. r . T h e q u e stio n w as th a t 'O u t of five best friends, how m any of them smoked?" + p< 10: * p < 0 5 : ** pc.O l: *** pc.OOl. X= 2.053. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 1 S.3.2.3 Interpersonal predictors Parental smoking significantly predicted adolescent smoking pro gression in that those with one parent who smoked had nearly twice like lihood to be slow escalators (OR = 1.58. p < .01). and those with two par ents who smoked had nearly three folds likelihood to be slow escalators (OR = 2.56, p < .001). Adolescents whose parents smoked had nearly five folds likelihood to be classified as stable smokers compared to those with no smoking parents (OR = 4.84, p < .05 for one smoking parent; OR = 4.44. p < .05 for two smoking parents). Friends' smoking exerted stronger influences on adolescent smok ing progression than parental smoking such that adolescents with one smoking friend had three folds risk to he classified as slow escalators (OR = 3.19. p < .001) and greater than four folds risk to be classified as stable smokers (OR = 4.69. p < .10). For adolescents with two or three smoking friends, the risk was even higher for being slow escalators (OR = 3.57, p < .001) or stable smokers (OR = 28.13. p < .001). Furthermore, adoles cents who had four or five smoking friends were nearly four times more likely to be slow escalators (OR = 3.63. p < .05) and had much larger risk to be stable smokers (OR = 161.43. p < .001). Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 2 5.3.3 Comparison between SAS PROC TRAJ and Mplus To validate the results from the SAS PROC TRAJ procedure, an identical model was established using the Mplus program, one specialized statistical package for latent growth mixture modeling. The results showed that when the latent growth mixture model was constructed in the SAS PROC TRAJ procedure and the Mplus with identical parame- terrzations, the parameter estimates of mean intercepts and slopes, and latent class proportions from both statistical programs was virtually the same (Table 5.14). The SAS PROC TRAJ is able to model both censored and uncensored normal models, while the Mplus is built with uncensored normal model. Both programs handle missing data using the maximum likelihood estimator with assumption of data missing at random (MAR). However, different scales of Bayesian Information Criteria (BIC) and Akaike Information Criteria (AIC) were used in the two programs. They were related to each other with a constant of - 2. i.e.. the BIC in Mplus approximately equals the BIC in SAS PROC TRAJ multiplied by - 2, for example. 8488.20 (Mplus) = (-2) ■ : (-4054.10) (Proc Traj). In addi tion. the Mplus program has a maximum likelihood estimator with robust standard errors, which is the default for the latent growth mixture mod eling. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 3 Table 5.14 C om p ariso n of p a ra m e te r e stim ate s an d model fit b etw een the SAS PROC TRAJ p ro c e d u re and the Mplus p ro g ram Parameter PROC TRAJ PROC TRAJ Mplus Estimate S.E. Estimate S.E. Estimate S.E. Model Estimator t Missing data ± Censored normal ML MAR t'neensored ML MAR normal I’ncensored normal MLR MAR Class 1 Intercept - 1.79 329.08 1.00*** 0.01 1.00*** 0.001 Linear 0.05 010.09 0.02 0.02 0.02*** 0.002 Quadratic - 0.07 SO 1.09 0.03 0.02 0.03*** 0.00-1 Class 2 Intercept 1.87*** 0.05 2.05*** 0.02 2.05*** 0.02 Linear 0.98*** 0.09 0.00*** 0.03 0.61*** 0.05 Quadratic - 0.58*** 0.1 1 - 0.55*** 0.0-1 - 0.35*** 0.06 Class 3 Intercept 1.50*** 0.2 1 1.25*** 0.05 ■1.25’ ** 0.12 Linear 1.5 1* 0.02 1.07*** 0.08 1.68** 0.35 Quadratic - 2.01** 0.90 - 2.05*** 0.12 - 2.65” 0. 1 1 Latent classes Proportion ("») C#1 09.2 -- 73.2 73.2 -- C#2 28. 1 - 23.7 23.7 - - c m 2.5 - - 3.1 3.1 - - Mean C*1 0.00 - - 0.00 - - C#2 - 0.89*** 0.05 - 1.13*** 0.05 -1.13*** 0.05 c m - 3.38*** 0.21 - 3.15*** 0.13 - 3.15*** 0.13 Model fit § BIC - 3735.58 - - 105 1.10 8-188.20 - AIC - 5701.82 - - 1020.3-1 8137.56 - Note: t. M L = m a x in u im likelihood: M L R = m a x im u m likelihood w ith robust s ta n d a r d errors. ±. M A R = m issing a t random . M issing d a ta w ere included for tim e -d e p e n d e n t variables, b u t excluded for tim e-stable c o v a ria te s for both Proc Traj a n d M plus. §. T he fo rm u la for BIC and AIC used in M p lu s w ere: BIC = - 2 log L + r In n: AIC = - 2 log L + 2 r (w here r is th e n u m b e r o f free m odel p ara m ete rs, n is th e sa m p le size). T he relation of BIC a n d A IC in SAS Proc Traj an d M plus w as: B IC iM plus) = - 2 BIC (Proc Traj): A IC (M plus) = - 2 AIC (Proc Traj). * p < 0 5 : ** pc.O l: *** p < 0 0 1 . n=2.053. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 5.4 Mediating Process on Smoking Trajectories 1 2 4 To test whether the program effects of the tobacco prevention trial on reducing the levels and growth rates of lifetime smoking was mediated through changing the growth rates of intention to smoke, a multilevel parallel process analysis in latent growth curve framework was con ducted. The direct effects of the prevention program on the growth fac tors of intention to smoke (mediator) and lifetime smoking (outcome) were estimated at school level, while the effect of growth factors of inten tion to smoke on the growth factors of lifetime smoking was estimated at individual level as shown in Figure o.T. The overall model fit was adequate (CFI = .92). The prevention program (predictor) was associated with reduced mean levels at time 2 (the point at which the initial status was set) (o.[ = -.047, p< .001) and slower growth rate (u.± = -.049, p< .001) of intention to smoke (mediator). Furthermore, the prevention program was related to reduced mean level (ti = -.10. p < .001). and slower growth rate (tj = -.07, p < .001) of lifetime smoking (outcome). In addition, the mean level and the growth rate of lifetime smoking were significantly related to the mean level (Pi = 1.90, p < .001), and the growth rate (Pu = l.oo, p < .001) of intention to smoke. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 5 M ed -0.5 T rend M ediator l.eu-l Mediator I Intervention p ro g ram | (l= p ro g ra m . 0=control) T ren d Tobacco o b \= l,9 8 7 (with com plete data); C'FI = .92; SR.MR,^,,.,.,,, = .15, S R -'IR |b c i» (tm ~ F igure 5.7 M ultilevel m ediational model for te stin g the program ef fect on sm oking trajectories m ed iated th ro u g h changing the m ean level and growth rate of in te n tio n to smoke. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 6 Nearly 50% of the total effects of the smoking prevention program on adolescent lifetime smoking were mediated through reducing the mean level (a\ x Pi = -0.089, p < .001) and the growth rate (a -> x p.< = - 0.076, p < .001) of intention to smoke (see Table 5.15). Table 5.15 P ro g ram effects on the trajecto ries of lifetime sm oking m ediated th ro u g h ch an g in g in ten tio n to smoke M ean level (in tercep t factor)_______(Iro w th ra te (slope factor) Path P a r a m e te r es tim a te S.F,. ’ P a r a m e te r estim a te S.E. P re d icto r (program I to m e d iato r (in te n - 0.0-17*** tion to smoke) (a) M e d ia to r (intention to sm oke) to outcom e (lifetim e smoking) 1.899*** U3) P re d icto r (p ro g ram ) to outcom e (lifetime sm o k in g ) (direct ef - 0.100**’ fect. t') M e d ia te d effect (In d ire c t effect, a ■ (1) - 0.089*** Z -test 1 - 3.567 p value 0.000 1 P e rc e n t of m ediated effect i -17.16 t . F o rm u la for Z-test = a - (1 / S.E. -m - +. Percent of m ediated effect w as calc +axp)] • 100%. §. S.E. = s ta n d a rd error. *** p < .001. 0.013 -0.0-19*** 0.007 0.087 1.55*** 0.220 0.021 -0.072*** 0.006 0.025 -0.076*** 0.010 - 1.800 < .0001 51 . 0.1 lated by the following formula: \a-[\ /(r’ Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 7 5.5 Interaction between Friends’ Smoking a n d Refusal Self- efficacy on Smoking Trajectories Cross-sectional data showed that the lowest prevalence rate of life time smoking was observed among adolescents who had no smoking friends and had high refusal self-efficacy, while the highest prevalence rate of lifetime smoking was seen among adolescents who had three to five friends who smoked and had low refusal self-efficacy (Figure 5.8). 100 O No FS. High R S E - -4 - 1-2 FS. High R S E ■ 3-5 FS High R S E » — No FS. Low RSE »- -1-2 FS. Low RSE p — 3-5 FS. Low RSE o c 0 ) ra > C 3 w 0. 7th Grade, Fall 7th Grade. Spring 8th Grade, Spring Grade F igure 5.8 Prevalence ra te s of lifetime sm oking by th e levels of friends’ sm o k in g (FS) and refusal self-efficacy (RSE). Lifetime sm o k in g w a s defined as having sm o k ed p a r t o r w hole cigarette or more in a subject’s w hole life. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 2 8 To test whether increasing number of friends who smoked inter acted with the changes in levels and growth rages of refusal self-efficacy in relation to smoking trajectories over time, an interaction model in la tent growth curve framework was established. The overall model fit was statistically adequate (CFI = .95. SRMR - .057) (see Figure 5.9). X = 2.053; C FI = .‘ > 5 ; SRMR = .057. Figure 5.9 Latent g ro w th curve model for te stin g in teractio n s be tw een frien d s’ sm oking and refusal self-efficacy on lifetime smoking. X u - Xu = friends' s m o k in g at time 1 to 3; Xm - Xm = refusal self-efficacy at time 1 to 3; Xu Xji - XnX-s = interaction te rm s b e tw e e n X u - X u and Xm - Xji. Tobi - Tobi = lifetime sm oking at tim e 1 to 3. Leveli - Leveli = latent in tercept factors. Trench - T r e n d i = la ten t slope factors. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. The data showed that the average numbers in the middle of 7th grade and the increasing rate of friends who smoked over time were sig nificantly associated with both the mean level and the growth rate of life time smoking in the middle of 7th grade. Although there were no signifi cant direct effects of refusal self-efficacy on either the mean level or the growth rate of lifetime smoking after adjusted for friends' smoking, the interactions between friends' smoking and self-efficacy were statistically significant for hoth mean level and growth rate of lifetime smoking such that adolescents who had fewer friends who smoked or higher refusal self-efficacy had lower mean levels and slower progression of lifetime smoking over time. 5.6 Generalized Estimating Equations (GEE) Modeling of Regular and Addictive Smoking Behaviors To further examine potential risk factors that might he associated with increasing trends of regular and addictive smoking as demonstrated in Figure 5.2. a generalized estimating equations modeling approach was used. There were significant linear and quadratic time effects for regular smoking ((3 = .68. p < .001 for linear effect; (1 = -.29, p < .001 fro quadratic effect) and addictive smoking ([) = -0, p < .10 for linear effect: |) = -.52. p Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 0 < .05 for quadratic effect) (see Table 5.16). Males were at greater risk for regular smoking ([) - 1.06. p < .001) and addictive smoking ([1 = 1.08. p < .001) than females. The tobacco prevention program was associated with reduced probabilities of regular smoking (P = -.36. p < .05) and addictive smoking (P = -.35. p < .10). Intention to smoke was related to greater probabilities of regular smoking (p = 1.91. p < .001) and addictive smok ing (P - 1.73. p < .001). Table 5.16 CfEE models for reg u lar an d addictive sm oking over tim e P redictor R eg u lar s P m oking S.E. Addictiv _ P e sm o k in g S.E. M ain effects T im e ( - l= baseline. 0=tim e 2. I=tim e8) 0.68 **’ o .os 0.29+ 0.15 (le n d e r (0=female. l=ma!e) 1.06*** 0.26 1 .O S " 0.8 1 E th m c itv f()=non-\Vhite. l=\V hite 0.02 0.27 0.20 0.85 S E S f 0.01 0.05 -0.002 0.07 G roup (-l= control. l= program ) -0.86* 0.1-1 -0.85+ 0.19 In te n tio n to sm oke (0=no. l=yes) j 9 j *★* 0.25 1.78*** 0.28 R efusal self-efficacy (0=no. l=yes) -0.12 0.25 -0.01 0.81 N u m b e r of p a re n ts who smoke (0=none. l=one parent.2= tw o p a re n ts) 0..16+ 0.22 0.87+ 0.28 N u m b e r of friends who sm oke 1.08*** 0.21 1.09*** 0.2-1 (0=none. l = l~ 2 friends. 2=8~5 friends) In te ra c tio n effects P a r e n t use ■ friend use -0.11 0.19 -0.87+ 0.19 T im e- -0.29*** 0.09 -0.52* 0.21 G roup ■ tim e -0.10 0.09 -0.21 0.16 G ro u p • tim e- 0.008 0.09 0.02 0.21 X ote: t. S E S rep rese n ts socioeconomic s ta tu s , w hich wa s defined a s (100 - p e rc e n ta g e of s tu d e n ts receiving free m e a l)/10. + p<. 10: * p < 0 5 : ** p< 0 1 : *** p < 0 0 1 . n=2.058. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 1 Parental smoking was marginally associated with the greater risk of regular smoking (p = .36. p < .10) and addictive smoking (p = .37. p < .10). In contrast, friends' smoking was significantly associated with the greater risk of regular smoking (P = 1.08. p < .001) and addictive smoking (P = 1.09. p < .001). The interaction between parental and friends' smok ing on addictive smoking was marginally significant (f) = -.37. p < .10). Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 2 1 CHAPTER 6 DISCUSSION 6.1 Major Findings of the Present Research The objectives of this research were: (a) to examine the patterns of adolescent smoking progressions (transition probabilities among stages ancl growth trajectories) over three time points by gender, diverse ethnic populations, and the prevention program: (b) to identify potential psycho social and behavioral (intrapersonal and interpersonal) risk factors, and demographic predictors associated with different progressive patterns and in particular the uptake of regular/addictive smoking: and(c) to ex plore mediating and/or moderating mechanisms between psychosocial and behavioral risk factors on smoking trajectories. The results showed that in general there was a significant increase of prevalence rates for lifetime smoking (from 1G% to 30%) and current smoking (from 3.2% to 7.2%) from seventh grade to eighth grade. The prevalence rates of smoking were higher in males. Hispanic and African Americans, and those in the control group compared to their counter- Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 3 parts. A significant linear increasing trend was observed for smoking prevalence rates. Proportions of smoking stages changed over three time points in that percentages of non-smokers decreased and percentages of trier/experimenters and regular/addictive smokers increased. A group of subjects who started smoking but stopped at later time points was also observed at time 2 and 3. A significant linear increasing trend was also observed for prevalence rates of regular/addictive smokers. The results of the latent transition analysis showed that overall 89% of adolescents were non-smokers at the beginning of seventh grade. 82% were non-smokers in the middle of seventh grade, and 7(>% were non-smokers in the middle of eighth grade, indicating a rapid progression of smoking during the transition period of starting seventh grade. The adolescents who had tried or experimented smoking had higher probabili ties of progressing toward more advanced smoking stages including cur rent smoking, daily smoking, and even addictive smoking. Males, adoles cents in the control group, and Whites had higher risk of transiting from lower to higher stages of smoking than females, those in the program group, and non-Whites in general. Three classes of adolescents were identified based on their smoking trajectories: stable non-smokers (G9.2%), slow escalators (28.4%), and stable smokers (2.4%). The slow escalators progressed from trier or ex Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 4 perimenters toward current smokers in a slow rate. Early stable smokers started as current smokers, progressed to regular smokers rapidly, and slightly decreased their use of cigarettes thereafter. Female adolescents and those in the program group had higher percentages of stables non- smokers. and lower percentages of stable smokers than males and those in the control group. Different ethnic groups of adolescents varied in their smoking trajectories. In general. Asians had highest percentages of stable non-smokers, followed by Whites. Hispanics had highest percent ages of slow escalators, followed by African Americans. African Ameri cans had highest percentage of stable smokers, followed by Whites and Asians. The results of multivariate logistic regression analysis confirmed that males. African Americans, adolescents with lower academic grades, those showed intention to smoke, those had smoking parents, and those had smoking friends were at greater risk to be classified as slow escala tors and/or stable smokers. In contrast, adolescents who were in the pro gram group, had higher socioeconomic status, and possessed refusal self- efficacy were at lower risk to be classified as slow escalators or stable smokers. The multilevel growth curve mediational analysis showed that nearly half of the total program effects of the tobacco prevention trial on the reduced levels and declined growth rates of lifetime smoking were Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 5 mediated through altering mean levels and growth rates of intention to smoke. Friends’ smoking interacted with refusal self-efficacy on the mean levels and growth rates of lifetime smoking in that adolescents who had higher refusal self-efficacy were less influenced by friends’ smoking on the mean levels and progression rates of their own smoking. The results of generalized estimating equations models revealed significant linear and quadratic time effects on adolescent regular and addictive smoking. In addition, males, those who showed intention to smoke, and those who had smoking friends were at greater risk of pro gressing to regular and addictive smoking. The program group was asso ciated with lower risk of regular and addictive smoking over time. 6.2 Comparison of Smoking Prevalence Rales with National, Florida, and California Sam ples Compared to the prevalence rates of current cigarette use as de fined by having smoked cigarette on at least one day of the 30 days pre ceding the survey in a national sample of public middle school stu dents'-0'’. and a large sample in Florida-0, the prevalence rate of current cigarette smoking in our sample was lower in general. For example, in the National Youth Tobacco Survey (n=15.058. N=131 schools) in 1999, Reproduced w ith perm ission of the copyright ow ner. Further reproduction prohibited w ithout perm ission. 13G the overall prevalence of cigarette use in middle school (grades 6 - 8) was 9.2%. Cigarette smoking rates were similar among boys (9.6%) and girls (8.8%), and among Whites (8.8%), African Americans (9.0%). and Hispan- ics (11.0%). In comparison, the Florida Youth Tobacco Survey (n=10.268. N=242 schools) in 1999 showed a higher prevalence rate of current ciga rette use among seventh graders (16.6%) and eighth graders (19.5%). The rates of current cigarette use were similar among males (15.0%) and female (14.9%). However, non-Hispanic black had lower current smoking rate (8.5%) than non-Hispanic White (16.1%) and Hispanics (16.1%). There was significant decline of current smoking rates among males and females, non-Hispanic White, and among seventh and eighth graders compared to the rates in 1998 in Florida. The prevalence rates of current cigarette use in our sample of middle school students (5.97% for 7'h grade and 7.14% for 8th grade) in Southern California were similar to the na tional sample but was lower than the Florida sample. In particular, the prevalence rate of lifetime smoking in African Americans was signifi cantly higher than that in Whites, and the prevalence rates of current and regular smoking in African Americans were similar to that in Whites in our research. This seemed to be inconsistent with previous findings showing that prevalence rates of smoking initiation and current smoking were lower in African American youths than White youths.-9-0'’ How Reproduced w ith perm ission of the copyright ow ner. Further reproduction prohibited w ithout perm ission. 1 3 7 ever, as shown in a recent national sample-0'', the prevalence rate of cur rent cigarette use was similar between African Americans (9.0%) and Whites (8.8%) in middle school, but this rate was lower in African Ameri cans (15.8%) than in Whites (32.8%) in high school. A caution needs to be given to generalize our findings to whole African American adolescents because of the small proportion in the entire sample (2.4%). The findings suggest that the prevalence rate of smoking among African American adolescents should be revisited in conjunction with how much population of African American students are represented at a school. The prevalence rates of lifetime and current (past 30 days) smok ing at eighth grade in the control group of our sample (n=990 in 19 schools in Orange County of Southern California) were lower than a rep resentative sample of eighth grade California youth (n=5.870 in G8 schools of 18 California counties during 1996-1997).-°1 There might he two possible explanations for the discrepancies of smoking prevalence rates between our sample and the representative California sample. First, because the prevalence rates of the California sample reflected the smoking status of adolescents at eighth grade in 1996-1997, our sample may represent a declining trend of adolescent smoking in California, as suggested by the 2000 National Household Survey on Drug Abuse.lso Second, the prevalence rates of current smoking in the control group of Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 8 our data were similar to the average prevalence rates of the national sample in 1999. The higher prevalence rates of smoking among eighth graders in the total California sample reflect that recent smoking rates among adolescents in Southern California might be lower in general compared to other regions in California. 6.3 Smoking Transitions during Early Adolescence Consistent to previous research, the transition patterns of smoking in this study indicated that the probabilities of adolescents who have al ready started trying or experimenting smoking at a prior time were more likely to progress to more advanced stages of smoking than never- smokers.7 7 ’1 ()i However this transition probability was still smaller than older adolescents7'110 1 or adults.1 !li In contrast to adolescents in New Zea land7 7 ’ where a largely one-way progression of smoking was found, we found reasonably large proportion of adolescents who were at higher smoking stages (current, daily, or addictive smoking) regressed to lower stages. This regression back to lower stages was especially true among adolescents who were in prevention groups, females, and non-Whites. The regression of smoking stages such as that addictive smoking goes down even in the control schools may reflect that the smoking de Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 3 9 creased in general in California as a result of tobacco education and pol icy efforts. Perhaps addiction in 8th grade adolescents is not as en trenched as addictive smoking in adulthood. In addition, it may he re sulted in more serious enforcement as a deterrent by 8th graders, when adolescents are preparing for graduation and transition to high school. These findings have two implications for smoking prevention pro grams. First, it clearly suggests the importance of preventing adolescents from making transitions from non-smokers to trier/experimenters, since the results have shown that once adolescents have started trying or ex perimenting smoking they progress toward more advanced levels of smoking rapidly. Second, it is equally important to prevent adolescents who have already started smoking from progressing to higher level of smoking, since the findings have clearly shown that these adolescent have great potential to reduce or stop smoking. 6.4 Trajectories and Risli Factors of Adolescent Smoking The resent study is the first attempt to examine the diverse growth trajectories of smoking among middle school adolescents, especially in dif ferent gender, ethnic groups, and prevention conditions. Our research provided further expansions and support to the recent s t u d i e s , 1, 21'> in Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 4 0 which five to six distinct growth trajectories of smoking were identified. It is important to identify the patterns and related risk factors during early adolescents because past studies have shown that during this pe riod of developmental process, adolescent may experience first rapid ini tiation and escalation of drug use including cigarette use.1 - ,y 1 1 Different developmental trajectories may be linked to distinct psycho-socio- behavioral antecedents. The results of our research were consistent to prior findings « » •* • < ;< > ” . so.**,in.r.um.-oo-iHun m that male gender, low socio-economic status, poor academic grades, intention to smoke, pa rental smoking, and friends' smoking were significant risk factors linking slow escalators and stable smokers. The tobacco policy and prevention program (Project TOPP) had significant effects in preventing adolescents from becoming slow escalators and stable smokers. After adjusted for all selected demographic, intrapersonal, and interpersonal risk factors, in tention to smoke and number of friends who smoked were the two strong est predictors for being slow escalators and stable smokers, suggesting adolescent’s decision to smoke and friends' influence play important roles for adolescent’s initiation and persisting use of cigarettes. The findings from the multilevel mediational analyses within the latent growth curve framework showed that both the reduced mean levels and decreasing rate of intention to smoke mediated near 50% of the effect of the tobacco pre Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 4 1 vention program on the levels and growth trajectories of lifetime smok ing, further suggesting that intention to smoke could be intervened as a mediator to reduce smoking levels and escalating rates. Refusal self-efficacy had a weak effect as a protector to adolescent smoking progression after adjusted for all other selected risk factors in the current data. Previous studies showed that low refusal self-efficacy predicted smoking initiation and escalation/1 "-1 -- Cigarette refusal among adolescent is a complex process and it may interact with other intraper sonal or interpersonal factors such as friends' smoking. The explanation for the lack of significant direct effects of refusal self-efficacy on smoking trajectories may be that the smoking resistance skills were not addressed in this program. However, as indicated in the findings in a latent growth curve moderator model in this study, adolescents with more smoking friends and lower refusal self-efficacy had higher prevalence rates and faster escalation of smoking. Friends' smoking significantly interacted with refusal self-efficacy in that adolescents who had higher refusal self- efficacy were less likely to he influenced by friends' smoking in terms of their own use of cigarettes, suggesting that school-based smoking preven tion programs aimed to increase adolescent’s cigarette refusal skills should be continued to be emphasized since it may operate as a buffer against friends’ influence. Friends’ smoking may also interact with pa Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 4 2 rental smoking in relation to adolescent use of cigarette and other drugs as shown in one of our previous studies in that adolescents who had both smoking parents and friends were at greater risk of using drugs including cigarettes.1 -1 6.5 Adolescent Nicotine Addiction and Kish Factors Our study reported the prevalence rates of addictive smoking be haviors among middle school adolescents of diverse ethnicity and exam ined the risk factors for addictive smoking behaviors for the first time. Compared to the prevalence of lifetime nicotine dependence in the US population aged 1” >-o4 years (24.1%),; and a nationally weighted sample of adolescents aged 12-17 years (3.8%)111. this rate was lower in middle school adolescents in our sample (2.92% for 7th graders and 2.83% for 8th graders). The criteria used in our study, i.e.. the composite scores of two items from Fagerstron Test for Nicotine Dependence measures.71:- !H were more liberal than the previous studies/’ 1 1 1 in which a diagnostic as sessment or a proxy measure of nicotine dependence were used. It is im portant to measure the prevalence rates of addictive smoking behaviors among adolescents because it may provide useful information about the extent or degree of adolescent nicotine dependence, and it may also be Reproduced w ith perm ission of the copyright ow ner. Further reproduction prohibited w ithout perm ission. 143 important to identify when and how adolescents progress to addictive smoking, and what factors may predict this progression. After adjusted for demographic, intrapersonal, and interpersonal factors, male gender, intention to smoke, and number of smoking friends were the strongest risk factors for adolescent addictive smoking behav iors, suggesting that addictive smoking behaviors during early adoles cents may primarily he determined by their own decision and friends' in fluence. Number of smoking friends seem to be as important for regular and addictive smoking as for initiation/trial smoking, which means that it is important to continue to consider friend smoking in early interven tion efforts as well. Addictive smoking behaviors progressed significantly over time, indicating the age effects on its developmental trend. In contrast, the to bacco policy and prevention program was associated with a lower likeli hood to addictive smoking, suggesting that the addictive smoking behav iors among early adolescents may be altered through effective interven tion programs. Parental influence on adolescent addictive smoking behavior was only marginally significant after taking other factors into account, sug gesting parents' role modeling exerts less impact on the advanced smok ing in adolescents. Although increasing evidence has shown the linkage Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 4 4 between genetic predispositions and addictive behaviors in family and twin studies,00 1!S! 1 1 it is not clear whether intention to smoke or the vul nerability to the environmental influences (e.g.. friends' smoking) is also genetically predisposed. The overall lack of effect of socioeconomic status (SES) and ethnicity on regular and addictive smoking suggests that our prevention efforts should continue to be universal, rather than segmented out for different ethnic groups or socioeconomic status. 6.6 Limitations and Strengths There were several limitations in this study. First, with three time points of data in growth curve modeling, the analytic power may be re duced and limited patterns of growth trajectories may be obtained as well. However, the large sample size (n=2.053) in our data may provide adequate compensation in terms of statistical power. For example, ac cording to Muthen & Curran,11 7 it only requires a sample size of 950 for a power of .80 to detect a small effect size (ES = .20) in a latent growth curve model with three waves of data. A power of greater than .90 can be achieved with our large sample size. Another advantage of our data set was that it comprised diverse ethnic populations, which allowed us to ex amine the patterns of smoking trajectories in such diverse populations. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 4 5 Second, all the measures were based on adolescent's self report, no biological measures or other source of data such as parental survey or friends’ survey were available to conduct cross-validation. However, sev eral past studies have already shown that perceived and actual use of others is highly correlated, and furthermore, that perceived use of others represents perceived social norms for use. a stronger predictor of adult drug use than actual use of others.ss 1U -11 It is especially necessary to validate the measures of addictive smoking with biomedical biomarkers. Fortunately, several previous studies showed that Fagerstron Test for Nicotine Dependence (FTXD) scores were highly correlated with saliva cotinine levels and thus were recommended as a valid measure for ado lescent smokers.I- 17’ ’ Third, participation attrition may be a problem given an overall at trition rate of 21.1% across three time points in the present study. As Cook and Campbell'’- have pointed out. attrition poses a threat to exter nal validity (generalizability) as well as to the internal validity of a study. Some researchers have argued that many prevention programs have shown apparently beneficial results due to participant attrition.1 8 This is unlikely the case in the present study because the attrition analy ses have shown that the likelihood of dropping out from the study among smokers at baseline was similar in the control group and program group Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 4 6 at time 2 and 3. Consistent to previous studies,87187 we found that boys, smokers at baseline, adolescent with low socio-economic status, and ado lescents in control group were more likely to dropout from the prevention program. The program effects and the associations between smoking progression and risk factors may vary between subjects who completed all three waves of surveys and those who did not.8187 Thus the gener- alizability of our findings may be limited to those adolescents who com pleted all three waves surveys. Fortunately, because there was no differ ential attrition between the prevention groups and the smoking outcome variables in our data, and all missing data were properly handled by an EM algorism"’181 ‘-8 1 8 5 or a maximum-likelihood estimating method in la tent growth mixture modeling,1 1 ,7 1 latent transition analyses,1 " ’ and generalized estimating equations modeling,1 -7 any threats to internal va lidity of our findings were very unlikely in the present study. Despite these limitations, the present research has the following strengths. First, we have used new advanced statistical methods to ex amine the patterns and related risk factors of adolescent smoking pro gression. and we have evaluated their validity and efficacy in application of smoking etiology and prevention. These new statistical methods are capable to control measurement errors and account for intraclass correla tions or design effects. Second, our research was the first study to iden Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. tify different pathways of smoking progression in junior high school ado lescents. especially in diverse ethnic populations in California. Third, we provided first empirical evidence of addictive smoking at early adoles cence. In summary, the finding of this research enhanced our under standing of gender and ethnic differences of smoking progression and re lated risk factors in California adolescents. It suggests that adolescents during the critical transition period of the first year of junior high school may experience a more rapid progression of smoking. The reinforcement of school nonsmoking policy may reduce the probability of initiating smoking and progressing toward higher level of smoking or may alter the pathw ays of adolescent smoking progression. Psychosocial and behav ioral variables, in particular intention to smoke and friends' smoking, should be efficiently targeted in order to prevent adolescent from initia tion and progression of cigarette smoking. 6.7 Implications for Prevention, Treatment, and Policy The findings of the present research have consistently shown that the Tobacco Program and Policy Trial (TOPP) aimed to prevent adoles cent from smoking had significant effect on low level smoking and more regular and addictive smoking as well. Most tobacco control programs Reproduced w ith perm ission of the copyright ow ner. Further reproduction prohibited w ithout perm ission. 148 implemented in the United States and aimed at youth have focused on "no-use" or "zero tolerance" prevention programs, which have shown net reductions of 20-G0% or more in smoking prevalence for up to 3 years post-program1 < il. However, the maintenance of effects in smoking inter vention programs is limited. As indicated by Botvin et al.-1 (pp. 1106), "Drug abuse prevention programs conducted during junior high school can produce meaningful and durable reductions in tobacco, alcohol, and m arijuana use if they (1) teach a combination of social resistance skills and general life skills. (2) are properly implemented, and (3) include at least two years of booster sessions." However, the effects of smoking in tervention programs decay over time. Regular implementation of a no use tobacco policy may be an effective complement to prevention pro gramming. A policy affects the entire population within its operational boundaries, for instance a school or a community, and it may reinforce prevention efforts by encouraging an anti-tobacco use social environ ment. I,is Furthermore, policies are more easily to change than social in fluences. psvcho-behavioral risk factors, and biogenetic factors. Unfortu nately. relatively few studies have yet evaluated the effectiveness of poli cies in reducing adolescent tobacco use. Studies conducted by Pentz and c oll ea gu ese va lua te d the smoking policy on the ban of smoking on Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 4 9 school grounds in middle and junior high in California. The results showed that school smoking policy was associated with reduced amounts of smoking rather than the prevalence rates, suggesting that current smoking policies for youth may work better for decreasing levels of smok ing than deterring smoking or promoting cessation among adolescents. A combined approach of social influences components with policy may maximize effects of tobacco prevention effects. Jason et al.mi showed a 40% reduction in smoking prevalence in the study of tobacco policy in conjunction with supportive mass media and school programming. In summary, the findings in our research suggest school policies and the enforcement of these policies have an impact on altering transi tions from lower level to advanced level of smoking, decreasing the risk of becoming a slow escalators and stable smokers, and even reducing the risk of regular and addictive smoking. In addition, reducing intention to smoke, decreasing friends’ and parent's influence, and increasing refusal self-efficacy should be emphasized in adolescents smoking prevention programs. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. REFERENCES 1 5 0 1. Abrams DB. Herzog TA. Emmons IvM. Linnan L. Stages of change versus addiction: a replication and extension. Nicotine & Tobacco Research. 2000:2(3):223-229. 2. Ajzen I. From decisions to actions: A theory of planned behavior. In: Kuhl J, Beckmann J. eds. Action-control: From cognition to be havior. Xew York: Springer: 1985:11-39. 3. Akaike H. Factor analysis and AIC. Psychometrikci. 1987:52:317- 332. 4. Andrews JA. Duncan SC. Examining the reciprocal relation be tween academic motivation and substance use: effects of family re lationships. self-esteem, and general deviance. • Journal of B ehav ioral Medicine. 1997:20(0): 523-5-19. 5. Andrews JA. Duncan SC. The effect of attitude on the development of adolescent cigarette use. -Journal of Substance Abuse. 1998: 10( 1): 1-7. 6. Anthony JC. W arner LA. Kessler RC. Comparative epidemiology of dependence on tobacco, alcohol, controlled substances, and in h a lants: basic findings from the national comorbidity survey. E x perimental Clinical Psychopharmacology. 1994:2:244-208. 7. Ary DY. Biglan A. Longitudinal changes in adolescent cigarette smoking behavior: onset and cessation. Journal of Behavioral Medicin e. 1988:11:301-382. 8. Ary DY, Biglan A. Longitudinal changes in adolescent cigarette smoking behavior: onset and cessation. Journal of Behavioral Medicin e. 1988:11 (4): 361 -382. 9. Bandura A. Socicd learning theory. Englewood Cliffs, XJ: Prentice Hall; 1977. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 151 10. Barnett T. O'Loughlin J. Paradis G. Renaud L. Reliability of proxy reports of parental smoking by elementary schoolchildren. Annals of Epidemiology. 1997:7(6):39G-399. 11. Baron RM. Kenny DA. The moderator-mediator distinction in so cial psychological research: conceptual, strategic, and statistical considerations. -Journal of Personality and Social Psychology. 1986: 51:1173-1182. 12. Bauman KE. Fisher LA, Bryan ES, Chenovveth RL. Antecedents, subjective expected utility, and behavior: a panel study of adoles cent cigarette smoking. Addictive Behaviors. 198-1:9:121-136. 13. Bentler PM. EQS structural equations program manual. Los Ange les: BMDP Statistical Software. Inc; 1989. 14. Bentler PM. Etiologies and consequences of adolescent drug use: implications for prevention. -Journal of Addictive Diseases. 1992: 11(3): 17-61. 15. Bentler PM. Bonett DG. Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin. 1980; 88(3):588-606. 16. Bentler PM. Chou CP. Practical issues in structural modeling. So ciological Methods and Research. 1987:16:78-117. 17. Biglan A. Duncan TE. Ary DV, Smolkowski K. Peer and parental influences on adolescent tobacco use. -Journal of Behavioral M edi cine. 1995:18(4):315-330. 18. Biglan A. Severson H. Ary D. Faller C, Gallison C. Thompson R, Glasgow R. Lichtenstein E. Do smoking prevention programs really work? Attrition and the internal and external validity of an evaluation of a refusal skills training program. -Journal of Behav ioral Medicine. 1987:10(2): 159-171. 19. Bollen KA. Total, direct and indirect effects in structural equation models. In: Clogg CC. ed. Sociological Methodology. Washington. DC: American Sociological Association: 1987:37-69. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 5 2 20. Bollen KA. Structural equations icith latent variables. New York: Wiley: 1989. 21. Botvin GJ. Baker E, Dusenbury L, Botvin EM. Diaz T. Long-term follow-up results of a randomized drug abuse prevention trial in a white middle-class population. JAMA. 1995:273(14): 1106-1112. 22. Breslau N. Peterson EL. Schultz LR. Chilcoat HD. Andreski P. Ma jor depression and stages of smoking. A longitudinal investigation. Archives of General Psychiatry. 1998:55(2):161-166. 23. Brook JS. Brook DW. Gordon AS. W hitem an M. Cohen P. The psy chosocial etiology of adolescent drug use: a family interactional ap proach. Genetic, Social, & General Psychology Monographs. 1990; 116(2): 111-267. 24. Bryk AS. Raudenbush SW. Hierarchical linear models: Applica tions and data analysis methods. Newbury Park. CA: Sage; 1992. 25. Burton P. Gurrin L. Sly P. Extending the simple linear regression model to account for correlated responses: an introduction to gen eralized estim ating equations and multi-level mixed modelling. Statistics in Medicine. 1998;17(11):1261-1291. 26. CDC. Smoking-attributable mortality and years of potential life of lost-United States. 1984. MMWR. 1997:46:444-451. 27. CDC. Cigarette smoking among high school students - LT nited States. 1997. MMWR. 1998;47:229-233. 28. CDC. Incidence of initiation of cigarette smoking - United States. 1965-1996. MMWR. 1998:47:837-840. 29. CDC. Tobacco use among middle and high school students - Flor ida. 1998 and 1999. MMWR. 1999:48( 12):248-253. 30. Charlton A. Minagawa KE, While D. Saying "no" to cigarettes: a reappraisal of adolescent refusal skills. Journal of Adolescence. 1999:22(5 ):695-707. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 5 3 31. Chassin L. Curran PJ. Hussong AM. Colder CR. The relation of parent alcoholism to adolescent substance use: a longitudinal fol low-up study. ■ Journal of Abnormal Psychology. 199G:105(l):70-80. 32. Chassin L. Presson CC. Pitts SC. Sherman SJ. The natural history of cigarette smoking from adolescence to adulthood in a midwest- ern community sample: multiple trajectories and their psychosocial correlates. Health Psychology. 2000:19(3):223-231. 33. Chassin L. Presson CC. Rose JS. Sherman SJ. The natural history of cigarette smoking from adolescence to adulthood: demographic predictors of continuity and change. Health Psychology. 1996:15(6): 478-4S4. 34. Chassin L. Presson CC. Sherm an SJ. Edwards DA. The natural history of cigarette smoking: predicting young-adult smoking out comes from adolescent smoking patterns. Health Psychology. 1990: 9(6):701-716. 35. Chassin L. Presson CC. Sherm an SJ. Edwards DA. Four pathw ays to young-adult smoking status: adolescent social-psychological a n tecedents in a nudwestern community sample. Health Psychology. 1991:10(6):409-418. 36. Chassin L. Presson CC. Sherm an SJ, Montello D. McGrevv J. Changes in peer and parent influence during adolescence: longitu dinal vs. cross-sectional perspectives on smoking initiation. Devel opmental Psychology. 1986:22:327-334. 37. Chassin L, Presson CC. Sherm an SJCE, Olshavsky R\V. Predicting the onset of cigarette smoking in adolescents: a longitudinal study. -Journal of applied social psychology. 1984:14:224-243. 38. Chen K. Kandel DB. The natural history of drug use from adoles cence to the mid-thirties in a general population sample. American Journal of Public Health. 1995:85(l):41-47. 39. Choi \VS, Pierce JP. Gilpin EA. Farkas AJ, Berry CC. Which ado lescent experimenters progress to established smoking in the United States. American -Journal of Preventive Medicine. 1997: 13(5):385-391. Reproduced w ith perm ission of the copyright ow ner. Further reproduction prohibited w ithout perm ission. 1 5 4 40. Chou CP. Bentler PM. Pent/. MA. Comparisons of two statistical approaches to study growth curves: the multilevel model and the latent curve analysis. Structural Equation Modeling. 1998:5:247- 26G. 41. Cohen S. Wills TA. Stress, social support, and the buffering hy pothesis. Psychological Bulletin. 1985:98(2):310-357. 42. Colby SM. Tiffany ST. Shiffman S. X iaura RS. Are adolescent smokers dependent on nicotine? A review of the evidence. Drug & Alcohol Dependence. 2000:59(Suppl l):S83-95. 43. Colby SM. Tiffany ST. Shiffman S. X iaura RS. Measuring nicotine dependence among youth: a review of available approaches and in strum ents. Drug & Alcohol Dependence. 2000:59(Suppl l):S23-39. 44. Colder CR. Mehta P. Balanda K. Campbell RT, Mayhew KP, Stanton WR. Pent/. MA. Flay BR. Identifying trajectories of adoles cent smoking: an application of latent growth mixture modeling. Health Psychology. 2001:20(2): 127-135. 45. Collins LM. Latent transition analysis for windows, cersion 2.3. University Park. PA: Methodology Center. Pennsylvania State University: 1998. 46. Collins LM. Flaherty BP. Hyatt SL, Shafer JL. WinLTA user's guide, version 2.0. University Park. PA: Methodology Center. Pennsylvania State University: 1998. 47. Collins LM. Graham J\V. Long JD. H ansen WB. Crossvalidation of latent class models of early substance use onset. Multivariate Be havioral Research. 1994:29(2): 165-183. 48. Collins LM, Hyatt SL. Graham JW. L atent transition analysis as a way of testing models of stage-sequential changes in longitudinal data. In: Little TD. Schnabel KLT , B aum ert J. eds. Modeling longi tudinal and multilevel data: practical issues, applied approaches, and specific examples. Mahwah, XJ: Lawrence Erlbaum: 2000:147- 161. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 155 49. Collins LM, Sussm an S. Rauch JM. Dent CW. Psychosocial predic tors of young adolescent cigarette smoking: a sixteen-month, three- wave longitudinal study. Journal of applied social psychology. 1987:17:554-573. 50. Collins LM. W ugalter SE. Latent class models for stage-sequential dynamic latent variables. Multivariate Behavioral Research. 1992: 27(1):131-157. 51. Conrad KM. Flay BR. Hill D. Why children start smoking ciga rettes: predictors of onset. British Journal of Addiction. 1992: 87(12):1711-1724. 52. Cook TD. Campbell DT. Quasi-experimentation: design & analysis issues for field settings. Chicago: Rand McNally: 1979. 53. Dempster AP. Laird XM, Rubin DB. M aximum Likelihood from Incomplete Data via the EM Algorithm. Journal of Royal Statisti cal Society. Series B (Methodological). 1977:39( 1): 1-38. 54. Dent CW. Sussm an S. Flay BR. The use of archival data to select and assign schools in a drug prevention trial. Evaluation Review. 1993:17(2): 159-181. 55. DiFran/.a JR. Rigotti NA. McNeill AD. Ockene JK. Savageau JA, St Cyr D. Coleman M. Initial symptoms of nicotine dependence in adolescents. Tobacco Control. 2000:9(3):313-319. 56. Diggle PJ. Liang KY. Zeger SL. Analysis of longitudinal data. Ox ford. England: Clarendon Press: 1994. 57. Dishion TJ. Capaldi DM. Yoerger K. Middle childhood antecedents to progression in male adolescent substance use: an ecological analysis of risk and protection. Journal of Adolescent Research. 1999:14:175-205. 58. Distefan JM. Gilpin EA. Choi WS, Pierce JP. Parental influences predict adolescent smoking in the United States, 1989-1993. Jour nal of Adolescent Health. 1998:22(6):466-474. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 5 6 59. Duncan SC. Duncan TE. Hops H. Progressions of alcohol, cigarette, and marijuana use in adolescence. -Journal of Behavioral Medicine. 1998:21:375-388. 60. Duncan SC. Duncan TE. Strycker LA. Risk and protective factors influencing adolescent problem behavior: a multivariate latent growth curve analysis. Annals of Behavioral Medicine. 2000:22(2): 103-100. 61. Duncan SC. Strycker LA. Duncan TE. Exploring associations in developmental trends of adolescent substance use and risky sexual behavior in a high-risk population. -Journal of Behavioral Medicine. l990:22(l):21-34. 62. Duncan TE. Duncan SC. Albert A. Li F. Strycker LA. An Introduc tion to Latent Variable Growth Curve Modeling: Concepts, Issues, and Applications. Mahwah. XJ: Lawrence Erlbaum Associates: 1900. 63. Duncan TE. Tildesley E, Duncan SC. Hops H. The consistency of family and peer influences on the development of substance use in adolescence. Addiction. 1995:90(12): 1047-1060. 64. Du •yer JH. Statistical models for the social and behavioral sciences. Xew York: Oxford University Press: 1083. 65. Dwyer JH. Feinleib M. Introduction to statistical models for longi tudinal observation. In: Dwyer JH. Feinleib M. Lippert P. Hoff- mcister H. eds. Statistical models for longitudinal studies of health. Xew York: Oxford University Press: 1002:3-47. 66. Dwyer JH. MacKinnon DP, Pentz MA, Flay BR, Hansen V YB, Wang EYI. Johnson CA. Estimating intervention effects in longi tudinal studies. American -Journal of Epidemiology. 1989,130:781- 795. 67. Ellickson PL. Bell RM. McGuigan K. Preventing adolescent drug use: long-term results of a junior high program. American -Journal of Public Health. 1993:83(6):85G-861. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 5 7 68. Ennet ST. Tobler XS. Ringwalt CL. Flewelling RL. How effective is drug abuse resistance education? A meta-analysis of project DARE outcome evaluations. American -Journal of Public Health. 199-1:84: 1394-1401. 69. Etter JF. Bergman MM. Hum air JP. Perneger TV. Development and validation of a scale measuring self-efficacy of current and former smokers. Addiction. 2000:95(6):901-913. 70. Fagerstrom KO. Measuring degree of physical dependence to to bacco smoking with reference to individualization of treatment. Addictive Behaviors. 1978;3(3-4):235-241. 71. Fagerstrom KO, Heatherton TF. Kozlowski LT. Xicotine addiction and its assessment. Ear, Xose, & Throat Journal. 1990:69(11):763- 765. 72. Fagerstrom KO. Schneider XG. Measuring nicotine dependence: a review of the Fagerstrom Tolerance Questionnaire. Journal of Be havioral Medicine. 1989:12(2): 159-182. 73. Farkas AJ. Pierce JP. Zhu SH, Rosbrook B. Gilpin EA, Berry C. Kaplan RM. Addiction versus stages of change models in predicting smoking cessation. Addiction. 1996:91(9):1271-1280: discussion 1281-1292. 74. Feldman HA. Families of lines: random effects in linear regression analysis. -Journal of Applied Physiology. 1988:64(1721-1732). 75. Fergusson DM, Horwood LJ. Transitions to cigarette smoking dur ing adolescence. Addictive Behaviors. 1995;20:627-642. 76. Flay BR. Youth tobacco use: risks, patterns, and control. In: Or leans CT. Slade J. eds. Xicotine Addiction: Principals and man agement. Xew York: Oxford LTiiversity Press: 1993:365-384. 77. Flay BR, Hu FB, Richardson J. Psychosocial predictors of different stages of cigarette smoking among high school students. Preventive Medicin e. 1998 ;27: A9-A18. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 158 78. Flay BR. Hu RB. Siddiqui 0, Day LE, Hedeker D. Petra it is J. Richardson J. Sussm an S. Differential influence of parental smok ing and friend's smoking on adolescent initiation and escalation of smoking. Journal of Health and Social Behavior. 1994:35:248-265. 79. Flay BR, Ryan KB. Best A, Brown KS. Kersell MW. d'Avernas JR. Zanna MP. Are social-psychological smoking prevention programs effective? The Waterloo Study. Journal of Behavioral Medicine. 1985:8:37-59. 80. Flint AJ. Yamada EG. Xovotny TE. Black-white differences in cigarette smoking uptake: progression from adolescent experimen tation to regular use. Preventive Medicine. 1998:127:358-364. 81. Goldstein H. Efficient statistical modelling of longitudinal data. Annals of Human Biology. 1986:13(2):129-141. 82. Goodman LA. Exploratory latent structure analysis using both identifiable and unidentifiable models. Biometrika. 1974:01:215- 231. 83. Graham JW. Collins LM. Wugalter SE. Chung NK. Hansen WB. Modeling transitions in latent stage-sequential processes: a sub stance use prevention example. Journal of Consulting < £ • Clinical Psychology. 1991;59( 1 ):48-57. 84. Graham JW. Donaldson SI. Evaluating interventions with differ ential attrition: the importance of nonresponse mechanisms and use of follow-up data. Journal of Applied Psychology. 1993:78(1): 119-128. 85. Graham JW. Flay BR. Johnson CA. Hansen WB, Grossman L, So- bel JL. 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. 1984; 14(2): 175-193. 86. Guo J, Collins LM, Hill KG. Hawkins JD. Developmental p ath ways to alcohol abuse and dependence in young adulthood. Journal of Studies on Alcohol. 2000:61(6):799-808. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 5 9 87. Hansen WB. Collins LM. Malotte CK. Johnson CA. Fielding JE. Attrition in prevention research. Journal of Behavioral Medicine. 1985:8(3): 261-275. 88. Hansen WB. Malotte CI\. Perceived personal immunity: the devel opment of beliefs about susceptibility to the consequences of smok ing. Preventive Medicine. 1980:15:3(53-372. 89. Hawkins JD. Catalano RF. Miller JY. Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: implications for substance abuse prevention. Psycho logical Bulletin. 1992:112( 1):0J-105. 90. Heath AC. Martin XG. Genetic models for the natural history of smoking: evidence for a genetic influence on smoking persistence. Addictive Behaviors. 1993; 18(1): 19-34. 91. Heatherton TF. Kozlowski LT. Frecker RC. Fagerstrom KO. The Fagerstrom Test for Xicotine Dependence: a revision of the Fager strom Tolerance Questionnaire. British Journal of Addiction. 1991: 86(9): 1119-1127. 92. Heatherton TF. Kozlowski LT. Frecker RC. Rickert W. Robinson J. Measuring the heaviness of smoking: using self-reported time to the first cigarette of the day and num ber of cigarettes smoked per day. British Journal of Addiction. 1989:84(7):791-799. 93. Hedeker D. A mixed-effects multinomial logistic regression model. Statistics in Medicine. In press. 94. Hedeker D. Gibbons RD. MIXOR: a computer program for nuxed- effects ordinal regression analysis. Computer Methods 6c Ptogrciins in Biomedicine. 1996:49(2):157-176. 95. Hedeker D. Gibbons RD. Flay BR. Random-effects regression mod els for clustered data with an example from smoking prevention research. Journal of Consulting & Clinical Psychology. 1994:62(4): 757-765. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 0 96. Hedeker D. Mormelstcin RJ. Analysis of longitudinal substance use outcomes using ordinal random-effccts regression models. Addiction. 2000:95(Suppl 3):S381-394. 97. Hoffmann JP, Cerbone FG. Su SS. A growth curve analysis of stress and adolescent drug use. Substance Use and Misuse. 2000: 35:687-716. 98. Holmbeck GN. Toward terminological, conceptual, and statistical clarity in the study of mediators and moderators: examples from the child-clinical and pediatric psychology literatures. Journal of Consulting & Clinical Psychology. 1997:65(4):599-610. 99. Hu FB. Flay BR. Hedeker DSO. Day LE. The influences of friends’ and parental smoking on adolescent smoking behavior: the effects of time and prior smoking. Journal of Social and Psychology. 1995: 25:2018-2047. 100. Hu LT. Bentler PM. Cutoff criteria for fit indices in covariance structure analysis: conventional criteria versus new alternatives. Structured Equation Modeling. 1999:6:1-55. 101. Jackson C. Initial and experimental stages of tobacco and alcohol use during late childhood: relation to peer, parent, and personal risk factors. Addictive Behaviors. 1997:22:685-698. 102. Jam es LR. Brett JM. Mediators, moderators, and tests for media tion. Journal of Applied Psychology. 1984:69(2):307-321. 103. Janson H. Longitudinal patterns of tobacco smoking from child hood to middle age. Addictive Behaviors. 1999:24:239-249. 104. J ason LA, Ji PY. Anes MD. Birkhead SH. Active enforcement of cigarette control laws in the prevention of cigarette sales to mi nors. JAMA. 1991:266:3159-3161. 105. Jessor R, & Jessor. S. L. Problem behavior and psychosocial devel opment. Xew York: Academic Press: 1977. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 1 106. Johnson CA. Pent/. MA. Weber MD. Dwyer JH. Baer X. MacKinnon DP. Hansen WB. Flay BR. Relative effectiveness of comprehensive community programming for drug abuse prevention with high-risk and low-risk adolescents. -Journal of Consulting & Clinical Psychology. 1990:58:447-456. 107. Jones BL. Nagin DS, Kathryn R. A SAS procedure based on mix ture models for estimating developmental trajectories. Sociological Methods & Research. 2001:29(3):374-393. 108. Joreskog KG. Modeling development: using covariance structure models in longitudinal research. European Child & Adolescent Psy chiatry. 1996:5(Suppl l):8-10. 109. Joreskog KG. Vang F. Xon-linear structural equation models: the Kenny-Judd model with interaction effects. In: Marcoulides GA. Schumacker RE. eds. Adcancecl structural equation modeling: is sues and techniques. Mahwah, XJ: Lawrence Erlbaum Associa tions: 1996:57-88. 110. Judd CM. Kenny DA. Process analysis: estim ating mediation in treatm ent evaluations. Evaluation Review. 19S1:5(5):602-619. 111. Kandel D. Chen K. Warner LA. Kessler RC. Grant B. Prevalence and demographic correlates of symptoms of last year dependence on alcohol, nicotine, marijuana and cocaine in the U.S. population. Drug & Alcohol Dependence. 1997:44( 1): 11-29. 112. Kandel DB. Chen K. Extent of smoking and nicotine dependence in the L'nited States: 1991-1993. Xicotine & Tobacco Research. 2000:2(3):263-274. 113. Kandel DB, Kessler RC. Margulies RZ. Antecedents of adolescent initiation into stages of drug use: A developmental analysis. In: Kandel DB. ed. Longitudinal research on drug use: Empirical fin d ings and issues. Washington, DC: Hemisphere: 1978:73-99. 114. Kassel JD. Are adolescent smokers addicted to nicotine? The suit ability of the nicotine dependence construct as applied to adoles cents. -Journal of Child & Adolescent Substance Abuse. 2000:9(4): 27-49. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 2 115. Kawakami X. Takai A, Takatsuka X. Shimizu H. Eysenck's per sonality and tobacco/nicotine dependence in male ever-smokers in Japan. Addictive Behaviors. 2000:25(4):585-591. 116. Kellam SG. Ensminger ME, Simon MB. Mental health in first grade and teenage drug, alcohol, and cigarette use. Drug and Alco hol Dependence. 1980:5:273-304. 117. Kenny DA. Judd CM. Estimating the nonlinear and interaction ef fects of latent variables. Psychological Bulletin. 1984:96:201-210. 118. Krohn MD. M assey JL. Skinner VVF. Lauer RM. Social bonding theory and adolescent cigarette smoking: a longitudinal analysis. -Journal of Health and Social Behavior. 1983:24:337-349. 119. Krull JL. MacKinnon DP. Multilevel mediation modeling in group- based intervention studies. Evaluation Review. 1999:23(4): ! 18-444. 120. Labouvie E\V, Pandina RJ. Johnson V. Developmental trajectories of substance use in adolescence: differences and predictions. Inter national -Journal of Behavioral Development. 1991:14:305-328. 121. Laird XM. Ware JH. Random effects models for longitudinal data. Biomet rics. 1982:38:963-974. 122. Lawrence L. Rubinson L. Self-efficacy as a predictor of smoking behavior in young adoleslents. Addictive Behaviors. 1986:11:367- 382. 123. Leventhal H. Cleary PD. The smoking problem: a review of the re search and theory in behavioral risk modification. Psychological Bulletin. 1980:88:370-405. 124. Li C. Pentz MA. Chou C-P. Parents as gatekeepers of adolescent drug use risk: A moderator model. Addiction. LT nder review. 125. Li F, Duncan TE. Acock A. Modeling Interaction Effects in Latent Growth Curve Models. Structural Equation Modeling. 2000:7(4): 497-533. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 3 126. Li F. Duncan TE. Duncan SC. Yang-Wallentin F. Acock AC. Hops H. Interaction models with latent growth curves. In: Marcoulides GA, Schumacker RE. eds. Xew developments and techniques in structured equation modeling. Mahwah. NJ. USA: Lawrence Erl- baurn Associates. Inc.: 2001:173-201. 127. Liang KY. Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73:13-22. 128. Little RJ. Rubin DB. Statistical analysis with missing data. New York: John Wiley & Sons: 1987. 129. Little TD. Schnabel KU. Baumert J. eds. Modeling longitudinal and multilevel data: practical issues, applied approaches, and spe cific examples. Mahwah. NJ: Lawrence Erlbaum: 2000. 130. MacKinnon DP. Dwyer JH. Estim ating mediated effects in preven tion studies. Evaluation Review. 1993:17(2):144-1” > S . 131. MacKinnon DP. Johnson CA. Pentz MA. Dwyer JH. Hansen WB. Flay BR. Wang EY. Mediating mechanisms in a school-based drug prevention program: first-year effects of the Midwestern Preven tion Project. Health Psychology. 1991:10(3): 164-172. 132. MacKinnon DP. Johnson CA. Pentz MA, Dw yer JH. Hansen WB. Flay BR. Wang EYI. Mediating mechanism in a school-based drug prevention program: first-year effects of the Midwestern Preven tion Project. Health Psychology. 1991:10:164-172. 133. Martin RA. Yelicer WF. Fava JL. Latent transition analysis to the stages of change for smoking cessation. Addictive Behaviors. 1996:21(l):67-80. 134. Mayhew KP, Flay BR. Mott JA. Stages in the development of ado lescent smoking. Drug & Alcohol Dependence. 2000:59(Suppl 1 ):S61-81. 135. McArdle JJ, Epstein D. Latent growth curves within developmen tal structural equation models. Child Development. 1987;58(1):110- 133. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 4 136. McCarthy WJ. Zhou Y. Hser YI. Individual change amid stable smoking patterns in polydrug users over 3 years. Addictive Behav iors. 2001:26:143-149. 137. McGiinis JM. Foege \YH. Actual causes of death in the United States. -JAMA. 1993:270:2207-2212. 138. McGue M. Elkins I. Iacono \VG. Genetic and environmental influ ences on adolescent substance use and abuse. American ■ Journal of Medical Genetics. 2000:96(5):671-677. 139. McNeill AD. Jarvis MJ. Stapleton JA. Russell MAEJR. Gammage P. Gray EM. Prospective study of factors predicting uptake of smoking in adolescents. -Journal of Epidemiology & Community Health. 1989:43:72-78. 140. Meredith \V. Tisak J. Latent curve a n a ly s is . Psychometrika. 1990:.",5:107-122. 141. Merikangas KR. Stolar M. Stevens DE. Goulet J. Preisig MA. Fen ton B. Zhang H. O'Malley SS. Rounsaville BJ. Familial transm is sion of substance use disorders. Archives of General Psychiatry. 1998:55:973-979. 142. Murray DM. Rooney BL. Hannan PJ. Peterson AX’. Ary D \’, Biglan A. Botvin GJ. Evans RI. Flay BR. Futterm an R. Intraclass correla tion among common measures of adolescent smoking: estimates, correlates, and applications in smoking prevention studies. Ameri can -Journal of Epidemiology. 1994:140(11):1038-1050. 143. Murry M. Swan AX’. Johnson MRD, Bewley BR. Some factors asso ciated with increased risk of smoking by children. British Journal of Psychiatry. 1983:24:223-232. 144. Muthen B. Latent variable mixture modeling. In: Marcoulides GA. Schumacker RE, eds. AT eiv developments and techniques in struc tural equation modeling. Mahwah. NJ: Lawrence Erlbaum Asso ciation: 2001:1-33. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 5 145. Muthen B. Muthon LK. Integrating person-centered and variable- centered analyses: growth mixture modeling with latent trajectory classes. Alcoholism: Clinical & Experimental Research. 2000:24(6):882-891. 146. Muthen BO. Second-generation structural equation modeling with a combination of categorical and continuous latent variables: new opportunities for latent class-latent growth modeling. In: Collins LM. Saver AG. eds. AT eiv methods for the analysis of change. W ash ington. DC: American Psychological Association: 2001:291-322. 147. Muthen BO. Curran PJ. General longitudinal modeling of individ ual differences in experimental designs: a latent variable frame work for analysis and power estimation. Psychological Methods. 100 7: 2(4): 3 71 -402. 148. Muthen BO. Shedden K. Finite mixture modeling with mixture outcomes using the EM algorithm. Biometrics. 1999:55:403-469. 149. Muthen LK. Muthen BO. Mplus user's guide. Los Angeles. CA: Muthen & Muthen; 1999. 150. Nagin DS, Tremblay RE. Trajectories of boys’ physical aggression, opposition, and hyperactivity on the path to physically violent and nonviolent juvenile delinquency. Child Development. 1999:70(5):1181-1190. 151. Newcomb MD. Psychosocial predictors and consequences of drug use: a developmental perspective within a prospective study. -Jour nal of Addictive Diseases. 1997:16( 1): 51 -89. 152. Newcomb MD. Chou CP, Bentler PM. Huba GJ. Cognitive motiva tions for drug use among adolescents: longitudinal tests of gender differences and predictors of change in drug use. -Journal of Coun seling Psychology. 1988:35(4):420-438. 153. Newcomb MD, Felix-Ortiz M. Multiple protective and risk factors for drug use and abuse: cross-sectional and prospective findings. -Journal of Personality & Social Psychology. 1992:63(2):280-296. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 6 154. Newcomb MD. Huba GJ. Bentier PM. Mothers' influence on the drug use of their children: confirmatory tests of direct modeling and mediational theories. Developmental Psychology. 1983:19:714- 726. 155. Newcomb MD. McCarthy \VJ. Bentier PM. Cigarette smoking, academic lifestyle, and social impact efficacy: an eight-year study from early adolescence to young adulthood. -Journal of applied so cial psychology. 1989:19:251-281. 156. Norton EC. Bieler GS. Ennett ST, Zarkin GA. Analysis of preven tion program effectiveness with clustered data using generalized estim ating equations. -Journal of Consulting & Clinical Psychology. 1996:64(51:919-926. 157. Patton GC. Carlin JB. Coffey C. Hibbcrt M. Bowes G. The course of early smoking: a population-based cohort study over three years. Addiction. 1998:93:1251-1260. 158. Patton GC. Carlin JB. Coffey C, Wolfe R. Hihbert M. Bowes G. The course of early smoking: a , , ' '' n-basod cohort study over three years. Addiction. 1998:93(8): 1251-1260. 159. Pede rson LL. Lefcoe NM. Change in smoking status am ong a co hort of late adolescents: prediction and explanation of initiation, maintenance and cessation. International -Journal of Epidemiology. 1986:15:519-526. 160. P ent/ MA. Adaptive evaluation strategies for estim ating effects of community-based drug abuse prevention programs. -Journal of Community Psychology. 1994.CSAP Special Issue: 26-51. 161. Pentz MA. Local policy change as a prevention strategy: proposed research methods. In: Bukoski WJ, Amsel Z, eds. Drug abuse pre vention: sourcebook on strategies and research. Westport: Green wood Publishing: 1997. 162. Pentz MA. School Tobacco Program and Policy Tried: Technical Report. Los Angeles: Departm ent of Preventive Medicine. Univer sity of Southern California: 2000. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 5382 1 6 7 163. Pentz MA. Brannon BR. Charlin VL. Barrett EJ, MacKinnon DP. Flay BR. The power of policy: the relationship of smoking policy to adolescent smoking. American Journal of Public Health. 1989:79:857-862. 164. Pentz MA. Chou CP. Measurement invariance in longitudinal clinical research assuming change from development and interven tion. Journal of Consulting and Clinical Psychology. 1994:62:450- 462. 165. Pentz MA. Dwyer JH. MacKinnon DP. Flay BR. H ansen WB. Wang EYI. Johnson CA. A multicommunity trial for primary pre vention of adolescent drug abuse. -/.-Ll/A. 1989:261:3259-3266. 166. Pentz MA. Li C. The Gateway theory applied to prevention. In: Kandel DB. ed. Stages and Pathways of Drug Involvement: Exam ining the Gateway Hypothesis. Xew York: Cambridge L’niversity Press: 2001:139-157. 167. Pentz MA. MacKinnon DP. Dwyer JH, Wang EYI. Hansen WB. Flay BR. Johnson CA. Longitudinal effects of the Midwestern Pre vention Project on regular and experimental smoking in adoles cents. Preventive Medicine. 1989:18:304-321. 168. Pentz MA. Sussman S. Newman T. The conflict between least harm and no-use tobacco policy for youth: ethnical and policy im plications. Addiction. 1997:92:1165-1173. 169. Pierce JP. Choi WS. Gilpin EA. Farkas AJ. Arthur J. Berry CC. Tobacco industry promotion of cigarettes and adolescent smoking. JAMA. 1998:279:511-515. 170. Pierce JP, Choi WS, Gilpin EA. Farkas AJ. Arthur J, Berry CC. Tobacco industry promotion of cigarettes and adolescent smoking. JAMA. 1998:279:511-515. 171. Pie rce JP. Farkas AJ. Gilpin EA. Beyond stages of change: the quitting continuum measures progress towards successful smoking cessation. Addiction. 1998;93(2):277-286. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 8 172. Pierce JP. Gilpin EA, Emery SL, Farkas AJ. Zhu SH. Choi WS. Berry CC, Distefan JM, White MM, Soroko S, Navarro A. Tobacco control m California: Who's winning the war? An evaluation of the tobacco program, 19S9-1996. La Jolla. CA: University of California San Diego: 1998. 173. Pomerleau CS. Carton SM. Lutzke ML. Flessland KA. Pomerleau OF. Reliability of the Fagerstrom Tolerance Questionnaire and the Fagerstrom Test for Nicotine Dependence. Addictive Behaviors. 1994:19(l):33-39. 174. Prochaska JO. DiClemente CC. Stages and processes of self-change of smoking: toward an integrative model of change. Journal of Consulting & Clinical Psychology. 1983:51(3):390-395. 175. Prokhorov AV, De Moor C. Pallonen L'E. Hudmon KS, Koehly L. Hu S. Validation of the modified Fagerstrom tolerance question naire with salivary cotinine among adolescents. Addictive Behav iors. 2000:25(3):429-433. 176. Rehoussin BA. Liang KY. Rehoussin DM. Estimating equations for a latent transition model with multiple discrete indicators. Biomet rics. 1999:55(3):839-84 5. 177. Rehoussin BA. Rehoussin DM. Liang K-Y. Anthony JC. Latent transition modeling of progression of health-risk behavior. M ulti variate Behavioral Research. 1998:33(4):457-478. 178. Richardson JL. Radziszewska B, Dent CW, Flay BR. Relationship between after-school care of adolescents and substance use. risk taking, depressed mood, and academic achievement. Pediatrics. 1993:92:32-38. 179. Rojas NL, Killen JD, Haydel KF, Robinson TN. Nicotine depend ence among adolescent smokers. Archives of Pediatrics & Adoles cent Medicine. 1998:152(2):151-156. 180. SAMHSA. Sum m ary of Findings from the 2000 Xcitional House hold Survey on Drug Abuse. Rockville. MD: Office of Applied Stud ies. NHSDA Series H-13, DHHS Publication No. (SMA) 01-3549.: 2001. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 6 9 181. Sargent JD. Dalton M. Beach M, Bernhardt A. H eatherton T. Ste vens M. Effect of cigarette promotion on smoking uptake among adolescents. Preventive Medicine. 2000:30:320-327. 182. SAS Institute. S A S /C Online Doe TM, Release S.l. Cary. NC: SAS Institute Inc.: 2000. 183. Schafer JL. Analysis of incomplete multivariate data. London: Chapman & Hall: 1997. 184. Schwarz G. Estim ating the dimension of a model. Annals of Statis tics. 1978:6:461-464. 185. Scott AJ. Holt D. The effect of two-stage sampling on ordinary least squares methods. -Journal of the American Statisticcd Associa tion. 1982:77:848-854. 186. Shiffman S, Balbanis M. Associations between alcohol and tobacco. In: Fetig JB, Allen JP. eds. Alcohol and tobacco: from basic science to clinical practice. Bethesda. MD: XIH/XIAAA Publication Xo. 95- 3931: 1995:17-36. 187. Siddiqui O. Flay BR. Hu FB. Factors affecting attrition in a longi tudinal smoking prevention study. Preventive Medicine. 1996:25(5):554-560. 188. Siegel M, Biener L. The impact of an antismoking media campaign on progression to established smoking: results of a longitudinal youth study. American - Journal of Public Health. 2000:90(3):380- 386. 189. Simon TR. Susstnan S. Dent C\V, Stacy AW. Burton D. Flay BR. Predictors of misreporting cigarette smoking initiation among ado lescents. Evaluation Review. 1996:20(5):552-567. 190. Singer J. Using SAS PROC MIXED to fit multilevel models, hier archical models, and individual growth models. -Journal of Educa tional ancl Behavioral Statistics. 1998:24:323-355. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 7 0 191. Skara SX. Sussman S. Dent CW. Predicting regular cigarette use among continuation high school students. American -Journal of Health Behavior. 2001:25:147-150. 192. Snijders TA. Bosker RJ. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. Thousand Oaks: Sage Publications: 1999. 193. Sobel ME. Asymptotic confidence intervals for indirect effects in structural equation models. In: Leinhardt S. ed. Sociological M eth odology. San Francisco: Jossey-Bass; 1982:290-312. 194. Sobel ME. Some new results on indirect effects and their standard errors in covariance structure analysis. In: Tuma XB. ed. Socio logical Methodology. Washington. DC: American Sociological Asso ciation: 1980:159-180. 195. Sobel ME. Direct and indirect effects in linear structural equation models. Sociological Methods & Research. 1987:10:155-170. 196. Sobel ME. Effect analysis and causation in linear structural equa tion models. Psychometrika. 1990:55(3):495-515. 197. Stacy AW. Xewcomb MD. Bentier PM. Interactive and higher- order effects of social influences on drug use. -Journal of Health and Social Behavior. 1992:33:220-241. 198. St acy AW. Sussman S. Dent CW. Burton D. Flay BR. Moderators of peer social influence in adolescent smoking. Personality and So cial Psychology Bulletin. 1992:18:103-172. 199. Stern RA. Prochaska JO, Velicer WF, Elder JP. Stages of adoles cent cigarette smoking acquisition: m easurem ent and sample pro files. Addictive Behaviors. 1987:12(4):319-329. 200. Sussm an S, Dent C. Flay BR. H ansen WB, Johnson CA. Psychoso cial predictors of cigarette smoking onset by White. Black, His panic. and Asian Adolescents in Southern California. Morbidity and Mortality Weekly Report. 1987:36:1 ls-lOs. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 7 1 201. Sussm an S. Dent CW, Burton D. Stacy AW, Flay BR. Developing school-based tobacco use prevention and cessation programs. Thou sand Oaks: Sage Publications: 1995. 202. Unger JB. Chen X. The role of social networks and media receptiv ity in predicting age of smoking initiation: a proportional hazard model of risk and protective factors. Addictive Behaviors. 1999:24:371-381. 203. Unger JB. Palmer PH, Dent CW, Rohrbach LA. Johnson CA. E th nic differences in adolescent smoking prevalence in California: are multi-ethnic youth at higher risk'.’ Tobacco Control. 2000:9(Suppl 2):II9-14. 204. USDHHS. Reducing the health consequences of smoking: 25 years of progress - a report of the Surgeon General. Rockville. Maryland: Public Health Service DHHS publication no. (CDC) 89-8411: 1989. 205. USDHHS. Preventing tobacco us among young people: a report of the Surgeon General. Atlanta. Georgia: US Department of Health and H um an Services Public Health Service, CDC National Center for Chronic Disease Prevention, and Health Promotion Office on Smoking and Health: 1994. 206. l T SDHHS. Tobacco use among middle and high school students - United States. 1999. MMWR. 2000:49(3):49-53. 207. USDHHS. Cigarette smoking among adults - United States. 1999. MMWR. 200l:50(40):869-873. 208. Wang MQ. Fitzhugh EC. Eddy JM. Fu Q. Turner L. Social influ ences on adolescents' smoking progression: a longitudinal analysis. American Journal of Health Behavior. 1997:21:111-117. 209. Wang MQ, Fitzhugh EC, Green BL, Turner LW. Eddy JM, Wester- field C. Prospective social-psychological factors of adolescent smok ing progression. Journal of Adolescent Health. 1999:24:2-9. 210. West P. Isenberg M. Instrum ent development: the Mental Health- Related Self-Care Agency Scale. Archives of Psychiatric Nursing. 1997:11(3):12G-132. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 7 2 211. West P. Sweeting H. Ecob RS. Family and friends' influences on the uptake of regular smoking from mid-adolescence to early adulthood. Addiction. 1999:94:1397-1411. 212. Willett JB. Measuring change: what individual growth modeling buys you. In: Amsel E. Renninger KA. eds. Change an d develop ment: issues of theory, method, and application. M ahwah. New Jer sey: Lawrence Erlbaum Associates; 1997:213-243. 213. Willett JB. Singer JD. M artin NC. The design and analysis of lon gitudinal studies of development and psychopathology in context: statistical models and methodological recommendations. Develop ment & Psychopathology. 1998:10(2):395-426. 214. Wills TA. Cleary SD. The validity of self-reports of smoking: analy ses by race/cthnicity in a school sample of urban adolescents. American Journal of Public Health. 1997:87( 1):56-G1. 215. Wills TA. McNamara G. Vaccaro D. Hirky AE. Escalated substance use: a longitudinal grouping analysis from early to middle adoles cence. Journal of Abnormal Psychology. 1996:105: Hid-180. 216. Zeger SL. Liang KY. An overview of methods for the analysis of longitudinal data. Statistics in Medicine. 1992:11( 14-15): 1825- 1839. 217. Zeger SL, Liang K-Y. Models for longitudinal data: a generalized estim ating equation approach. Biometrics. 1988:44:1049-10G0. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. APPENDIX A 1 7 3 Percentages of Gender and Ethnicity by School District and School School School G ender (*n) Ethnicity ( 0 \ « J District C ode1 N Girl Boy White A-A* Hispanic Asian O thers A naheim 101 '' 109 50,16 ■19,51 23.85 0.00 56.88 10.09 9.17 102 1 19 51.G8 18.32 27,52 5.37 14,30 12.08 10.71 10.1 78 51.28 18.72 16.67 5.13 55.13 15.38 7.69 104 89 19. 11 50.5G 17.19 1,19 19.10 15.73 13,18 105 r 90 50.00 50.00 28,89 7.78 35.56 10.67 1111 10G i1 1 1G ■18.28 51.72 18.97 0,86 65.52 8.02 0.03 107' 115 11.35 55.G5 2.61 0.00 89.57 3,18 4,35 108 i' 95 17.37 52.G3 33.68 5.26 1 1.71 37,89 8,12 109 i' 170 58.82 ■11.18 10.00 1.12 1 1.12 28,82 12.9 1 Subtotal -- 1.011 50.71 19.26 27.00 3.56 13.22 16.72 9.50 Irvine 110 102 52.91 ■17.06 69.61 0.98 8.82 8,82 11.76 111 132 51.52 18,18 53.03 2.27 3.79 30,30 10.61 112i' 103 18.51 51,16 55,11 0.97 2.91 25.2 1 15,53 113' 95 -1G.32 53.68 15.26 5.26 8,12 21.05 20 00 11 1 i’ 118 13.22 56.78 18.31 0.85 6.78 38.98 5.08 1 15 i' 17 ■IG.81 53.19 65.96 0.00 2.13 14,89 17.02 Subtotal r > t ) 7 7 3 ,// 51.59 55.11 I . S I 5. 70 21. 79 12.56 Newport- Mesa 11G 85 55.29 ■ 1 1.71 80.00 1.18 7.06 5,88 5,88 117 i' 92 G5.22 3-1.78 33.70 1.09 33.70 13.04 18,18 118 i' 123 58.5 1 11,16 77.21 0.00 8.9-1 3.25 10.57 119' 1 15 52.11 17.59 26.90 0.00 61,83 3,15 4.83 Subtotal .. 1 1 5 5 7 . 3 0 1 2 . 7 0 5 2 . 3 6 0 . 1 5 3 1 . 9 1 5 . S I 9 . 1 1 Total - 2.053 51.49 4 8 .5 1 40.67 2.39 29.86 1 6 .7 1 1 0 .38 Note, f . P - program group; C - the control group. +. A-A - African Americans. Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. APPENDIX B 1 7 4 M e a s u r e s o f t o b a c c o u s e s , s o c ia l i n f l u e n c e , i n t r a p e r s o n a l a n d DEMOGRAPHIC CHARACTERISTICS OF SURVEYS AT THREE WAVES C o n stru ct M easures T l (7th. fall) T2 (7,h. spring) T3 (8“'. spring) Tobacco use How m any cigarettes have you smoked in your whole life? < l=none, J=part or all one cig., 5=5-10 cigs., 1=11-50 cigs, 5=1-5 packs, 6=more than 5 packs.) T1Q21 T2Q2d T.JQ23 How m any cigarettes have you smoked in the past 50 days ’ ( l=nnr.c, 5=part or all one cit;., 5=5-10 cigs., 1=1 1-50 cigs, 5=1 -5 packs, 6=more than 5 packs.) T1Q22 T2Q2I T3Q21 During the past month, on the days you smoked, how many cigarettes did you smoked per day? ( l=<li(l not sm oked, 5=less than I cig., 5=1 cig. 1=5-5 cigs, 5=6-10 cigs, 6=1 l-'JO cigs, 7=rnnrc th a n JO cigs. ) T IQ 23 T2Q2-") T:K}25 Do you think you will smoke a cigarette in the next few months'.’ ( l= de/initely not, J=not sure, 5=prnbably, -l=definitely yes) TlQ:i I T2Q.j.i TdQdd Do you use tobacco within the first 50 m inutes of waking'.’ i l=no, J=yes, 5=don V sm oke) TlQdO T2Q32 TdQ dl Do you feel you are addicted to tobacco'.’ ( l=don't use tobacco, J=no, 5=ves) TlQoO T2Q o7 TdQdd P a re n ta l m odeling Of tho two adults whu are the most im portant in your life, how many do you think use cigarettes'.’ il=none, J=1 a d u lt, 5=5 adults) T1Q 10 T 2 Q I 1 TdQ Id P a re n ta l a p p ro v a l 1 would get in trouble with my p arents if they knew I smoked. ( I=strongly d i s agree, J=ilisagree, 5=not sure, l=agree, 5 =stronply a g n ■ < ■ ) T lQ ir, T2Q52 T dQ lS Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. A p p e n d ix B (Continued) C onstruct M easures T1 (7th. fall) T2 (7th. spring) T3 (8th. spring) How m any times have you been in trou ble with parents or family for using to bacco last year? ( l= noru\ 2=1 time, 3=2 tim es) T IQ 29 T 2 Q 3 1 T.'3Q30 F riend m odeling Of your five friends, how m any smoke cigarettes? ( 1=0, 2=1,3=2, 1=3,5=1. 6=5) TIQ.36 T 2 Q I0 T'3Q'39 F riend approval How would your friends act toward you. if you smoke cigarettes? ( l=i ery friendly, 2=pretty frien d ly . 3=pretty u n friendly. -l=Lvry u n frien d ly ) T lQ dS T 2Q 12 T.'JQ 11 Social norm s Out of every 100 people your age. how m any do you think use cigarettes at least once a month? ( 1=0. 2=10, 3=20, 1=30. 5=10, 6=50, 7=60, 3=70, 9=30. 10=90.11 = 100) TiQ dr, T2Q:)f) TdQ.'IS GPA What grades do you usually get in school'.’ ( l= m ostly .As, 2= 1 s a n d B's, 3= m ostly B ’ s, l=B's a n d Os, 5=5lastly C's, 6= (’'s a n d D's, 7=Mnstly D ’ s, 3= I)’ s a n d F's, 9=mostly F's) Tiqi T 2Q I T.iCJ I School bonding In the last school year, how many times did you skip school'.’ ( l= nnne. 2=once, 3=2 times, 1=3-1 tim es, 5=5 or more tim es) TlQ:i T2Q.1 T:JQ3 In ten tio n to sm oke Do you think you will smoke a cigarette in the next few m onths? ( l= definite!y not. 2=not sure, 3=probul)ly. -I=definitely T lQ d l T2Q.i:J TdQd.'S Refusal self- efficacy W.S) If your best friend offered you ciga rettes. how hard would it be to refuse the offer'.’ ) I=i ery h a rd , 2=hard. 3=i'asy. ■l=i ery e a s y ) Tig.'JT T 2Q ! 1 T 'iQ 10 G ender ( I=boy, 2=girl) Tlgend T2gend Togend D em o er er er grap h ic Race (I = \\'hite, 0 = X o n -\\ hite) T 1 race T2race T3race variable School level socio-economic status % S tudents given free meal at school Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. APPENDIX C 176 L a t e n t T r a n s it io n A n a l y s is P r o g r a m a n d P a r a m e t e r E s t i m a t e s * The basic model (M^ with free parameter estimates for uncondi tional and transition probabilities between control and program group. P R O G R A M S T A R T E D : T h u O c t 2 5 2 0 : 2 2 : 3 6 2 0 0 1 * T O P P d a t a i' 1 . 2 4 % m i s s i n g ) * 3 t i m e p o i n t s * 2 s t a t i c l a t e n t c l a s s e s * 5 d y n a m i c l a t e n t s t a t u s e s I N F O R M A T I O i l A B O U T T H I S J O B : C O N T R O L D A T A S E T : C : \ C h a o y a n g L i l C h a o y a n g L T A ' T O P P ' t o b g r p m l . c n t D A T A A N A L Y Z E D I N T H I S R U N R E A D F R O M : C : ' • . C h a o y a n g L i 1 ' C h a o y a n g '• L T A T O P P . t c b g r p . d a t O U T P U T P R I N T E D T O : C : \ C h a o y a n g L i l . C h a o y a n g L T A T O P P t o b g r p m l . o u t S T A T I C L A T E N T V A R I A B L E Y E S N U M B E R O F L A T E N T C L A S S E S 2 N U M B E R O F M A N I F E S T I T E M S 1 D Y N A M I C L A T E N T V A R I A B L E Y E S N U M B E R O F L A T E N T S T A T U S E S 5 N U M B E R O F O C C A S I O N S O F M E A S U R E M E N T 3 N U M B E R O F M A N I F E S T I T E M S P E P O C C A S I O N 4 T Y P E O F P R O C E S S F I R S T - O P D E P . N U M B E R O F S U B J E C T S 2 0 5 3 N U M B E R O F U N I Q U E R E S P O N S E P A T T E R N S 1 0 1 M A X I M U M N U M B E R O F I T E R A T I O N S 5 0 0 0 C O N V E R G E N C E C R I T E R I O N . 0 0 0 0 1 0 0 0 0 0 0 0 0 0 M I S S I N G D A T A I N R E S P O N S E P A T T E R N S Y E S P R I N T R E S I D U A L S Y E S T H E F O L L O W I N G C O N S T R A I N T S H A V E B E E N S P E C I F I E D W H E R E 0 = F I X E D T O S T A R T V A L U E 1 = F R E E 2 O R G R E A T E R M E A N S C O N S T R A I N E D E Q U A L T O A N Y O T H E R P A R A M E T E R W I T H T H E S A M E D E S I G N A T I O N Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 7 7 L I T T L E P . K O P A R A M E T E R S L I T T L E P . K O S A R E P R O B A B I L I T I E S O F R E S P O N S E S T O I T E M S M E A S U R I N G T H E S T A T I C L A T E N T V A R I A B L E C O N D I T I O N A L O N L A T E N T C L A S S M E M B E R S H I P R E S P O N S E C A T E G O R Y 1 ■ v c : A 0 R M . T i i r : 1 o ; L C O N T R O L C P R O G R A M 0 R E S P O N S E C A T E G O R Y 2 P A P R 0 G 1 R r\ M C O N T R O L 0 P R O G R A M 0 B I G P . H O P A R A M E T E R C O N S T R A I N T S B I G R H O S A R E P R O B A B I L I T I E S O F R E S P O N S E T O I T E M S M E A S U R I N G T H E D Y N A M I C L A T E N T V A R I A B L E C O N D I T I O N A L O N L A T E N T S T A T U S , L A T E N T C L A S S , A N D T I M E B I G P . H O C O N S T R A I N T S F O R L A T E N T C L A S S " C O N T R O L " A T T I M E 1 R E S P O N S E C A T E G O R Y 1 L M M N ; D N F N I 0 0 0 A 0 T 0 F N V U E c D C : I r I ' G ; G j G i 1 N O N S M K 2 2 2 2 2 2 2 2 N O C O R R 2 2 2 2 2 2 2 M O N S M K n 2 2 2 n D A Y S M K n o 2 ° A D D S M K 2 2 2 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 R E S P O N S E CATEGORY 2 L Y M Y D Y F Y I E 0 E A E T E ; F s N S Y S N S : E C C D C I r L I G a G M O N S M K 6 6 6 6 M O C O P . R 6 6 6 6 6 M O N S M K 6 6 6 6 6 6 D A Y S M K 6 6 6 6 6 6 6 A D D S M K 6 6 6 6 6 6 6 6 B I G R H O C O N S T R A I N T S F O P L A T E N T C L A S S " C O N T R O L " A T T I M E R E S P O N S E C A T E G O R Y 1 L M M N D N F N I 0 0 0 A 0 T 0 F : n Y N E c C D C T I r i. G v_J M O N S M K N O C O R R n -> 2 2 -i M O N S M K n -> n D A Y S M K -> n 2 -* n A D D S M K 2 2 R E S P O N S E C A T E G O R Y 2 L Y M Y D Y F Y T ” 0 E A E T E F S N S Y S N S t E ; C c D i c | T : i i i T l G I G G N O N S M K 6 6 6 6 N O C O R R 6 6 6 6 6 M O N S M K 6 6 6 6 6 6 D A Y S M K 6 6 6 6 6 6 6 A D D S M K 6 6 6 6 6 6 6 6 B I G R H O C O N S T R A I N T S F O P L A T E N T C L A S S " C O N T R O L " A T T I M E Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 7 9 RESPONSE CATEGORY 1 ' L N M N 1 D N F N : I 0 0 0 A O T 0 F H Y N : E C C D c I 1 i ! i G i G i M O N S M K n -) 2 2 2 2 2 2 N O C O R R 2 2 2 2 2 2 2 M O N S M K 2 D A Y S M K 2 2 2 n A D D S M K n 2 2 2 R E S P O N S E C A T E G O R Y 2 L Y M Y D Y F i I E 0 E A E T E F 5 ‘ i S Y S N 3 £ n n I I I G G G M O N S M K 6 6 6 6 M O C O R P . 6 6 6 6 6 M O N S M K 6 6 6 6 6 6 D A Y S M K 6 6 6 6 6 6 6 A D D S M K 6 6 6 6 6 6 6 6 B I G R H O C O N S T R A I N T S F O R L A T E N T C L A S S " P R O G R A M " A T T I M E 1 R E S P O N S E C A T E G O R Y 1 L N M N d u : f rr I 0 0 0 A 0 T 0 F ' M V H E r c D r > r T ; i G G : G N O N S M K 1 2 2 22 2 2 2 2 NOCORR 2 2 22 2 2 MONSMK 2 n 2 2 DAYSMK 2 2 2 2 2 ADDSMK 2 2 -> Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 8 0 N O N S M K N O C O R R M O N S M K D A Y S M K A D D S M K M O N S M K N O C O R R M O N S M K D A Y S M K A D D S M K N O N S M K N O C O R R M O N S M K D A Y S M K A D D S M K R E S P O N S E CATEGORY 2 L Y M Y D Y F Y I E 0 E A E T E F S N S Y S M S J? C n D C I I I G G G a 6 6 6 6 S 6 6 6 6 5 5 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 B I G R H O C O N S T R A I N T S F O P L A T E N T C L A S S " P R O G R A M " A T T I M E 2 R E S P O N S E C A T E G O R Y 1 M N 0 0 C I G 2 2 -i -i D N A 0 F N T 0 M D R E S P O N S E C A T E G O R Y 2 L Y 1 M Y . D Y : F Y I E : 0 E , A E ! T S F S : N S Y S ; N S E C C D c ; i ; i I G G G 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 5 6 66 6 6 6 66 66 6 6 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 8 1 B I G R H O C O N S T R A I N T S F O P . L A T E N T C L A S S " P R O G R A M " A T T I R E S P O N S E C A T E G O R Y i 1 1 L N M N : D N ■ F M ' I 0 0 0 . A C T 0 F N Y M E ( 2 D r* I I T G r * g j 1 G N O N S M K 2 2 2 2 2 2 n -> N O C O R R 2 ■? ^ 2 2 o o M O N S M K 2 2 D A Y S M K 2 — — A D D S M K - -1 - - R E S P O N S E C A T E G G R : — 1 L Y M Y D Y F Y I E 0 E A E T E F S M 3 Y S N S E Q n D n I i I -j G N O N S M K 6 6 6 6 N O C O R R 6 6 6 6 6 M O N S M K 6 6 6 6 6 6 D A Y S M K 6 6 6 6 6 6 6 A D D S M K 6 6 6 6 6 6 6 6 C O N S T P A I N T S F O R G A M Y A P A R A M E T E R S G A M M A S A R E U N C O N D I T I O N A L F R 0 E A 5 1 L I T I E S O F M E M B E R S H I P I N E A C H L A T E N T C L A S S O F T H E S T A T I C L A T E N T V A R I A B L E C O N T R O L X P R O G R A M 1 C O N S T R A I N T S F O R D E L T A P A R A M E T E R S D E L T A S A P E P R O B A B I L I T I E S O F L A T E N T S T A T U S M E M B E R S H I P C O N D I T I O N A L O N L A T E N T C L A S S D E L T A P A R A M E T E R C O N S T R A I N T S F O R L A T E N T C L A S S " C O N T R O L " N O N S M K N O C O R R M O N S M K D A Y S M K A D D S M K T I M E I 1 1 1 1 1 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 8 2 DE LTA PARAMETER C O N ST R A IN T S FOR LATENT CLASS "PROGRAM " T IM E 1 M O N S M K 1 N O C O R R 1 M O N S M K 1 D A Y S M K 1 A D D S M K 1 C O N S T R A I N T S F O P . T A U P A R A M E T E R S T A U S A P . E P R O B A B I L I T I E S O F L A T E N T S T A T U S M E M B E R S H I P A T T I M E T + l ( C O L U M N S ) C O N D I T I O N A L O N L A T E N T S T A T U S M E M B E R S H I P A T T I M E T ( R O W S ' A N D O N L A T E N T C L A S S M E M B E R S H I P T R A N S I T I O N P R O B A B I L I T I E S F O R L A T E N T C L A S S " C O N T R O L " R O W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 1 C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 2 N M M D A 0 0 ro D w U V D S 0 S S S M ? M M M K p Y . K K N O N S M K 1 X 1 1 i N O C O R R 0 1 1 1 T_ M O N S M K 0 1 1 i 1 D A Y S M K 0 1 1 1 A D D S M K 0 1 x i 1 R O W S R E P R E S E N T L A T E N T S T A T U S M E M 3 E P . S H I P A T T I M E 2 C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 3 N N ! M d : A 0 0 : 0 A D N C n ; V n S 0 s s s M p M M M K p . K K K i N O N S M K 1 1 1 1 1 N O C O R R 0 1 1 1 1 M O N S M K 0 1 1 1 1 D A Y S M K 0 I 1 1 i. A D D S M K 0 1 1 i T Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 8 3 T R A N S I T I O N P R O B A B I L I T I E S F O R L A T E N T C L A S S " P R O G R A M " R O W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 1 C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 2 N N M D A O O O A D N c ; N ! Y D S : o : S ; S ; S M ; P- ! M j M i M K P. : K : K K N O N S M K 1 1 1 1 N O C O R R 'J 1 X _ M O N S M K 0 .L 1 1 D A Y S M K 0 1 1 1 A D D S M K 0 1 1 1 R O W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 2 C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 3 N N M D A O O O D N c M Y D S i 0 S s | S M R M M : M K P. K K K N O N S M K 1 1 x x x N O C O R R 0 1 1 x 1 M O N S M K 0 1 1 x 1 D A Y S M K 0 1 1 1 1 A D D S M K 0 I 1 x x S T A R T V A L U E S L I T T L E R H O P A R A M E T E R S L I T T L E R H O S A R E P R O B A B I L I T I E S O F R E S P O N S E S T O I T E M S M E A S U R I N G T H E S T A T I C L A T E N T V A R I A B L E C O N D I T I O N A L O N L A T E N T C L A S S M E M B E R S H I P R E S P O N S E C A T E G O R Y 1 I V C I A O j ; P. N T i | 1 R | ! o | i L , C O N T R O L 1 . 0 0 0 P R O G R A M 0 . 0 0 0 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. R E S P O N S E CATEGORY 2 i V P A P. P. 0 G 1 P. M C O N T R O L 0 . 0 0 0 P R O G R A M 1 . 0 0 0 B I G P . H O P A R A M E T E R S B I G P . H O S A R E P R O B A B I L I T I E S O F R E S P O N S E S T O I T E M S M E A S U R I N G T H E D Y N A M I C L A T E N T V A R I A B L E C O N D I T I O N A L O N L A T E N T S T A T U S , L A T E N T C L A S S , A N D T I M E R H O P A R A M E T E R S F O R L A T E N T C L A S S " C O N T R O L " A T T I M R E S P O N S E C A T E G O R Y L N M t l D N F N I O 0 O A O T O F N Y N E C C D C I I I G G G N O N S M K 0 . 3 0 0 0 . 3 0 0 0 . 3 0 C 0 . 8 0 0 N O C O R R 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 M O N S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 D A Y S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 A D D S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 R E S P O N S E C A ' T E G O R Y 2 1 L Y M Y D Y F Y ; I E O E A E T E F S N S v S N S i E C c i D ! C I T i ; G C- ! G ! I N O N S M K 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 N O C O R R 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 M O N S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 D A Y S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 A D D S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. RHO PAR A M ETER S FOR LA TE N T C L A S S "CONTROL " AT T IM E 2 RESPONSE CATEGORY 1 | L N ! M N j D N j F N ! I o : o o j A O | T O ! F ; M ! Y ! N ! E c C D c [ ' I X G G ; ; g 1 NONSMK 0 . 300 0.800 0.800 0 . 8 0 0 NOCORR 0 . 050 0.050 0.050 0 .0 5 0 MONSMK 0.050 0.050 0.050 0 . 0 5 0 DAYSME 0.050 0.050 0.050 0 . 050 ADDSMK 0.050 0.050 0.050 0.0 50 RESPONSE CATEGORY - L Y M Y n y F Y I E O E A E T E F S N S Y S N S : E C C D c I I .L G G G NONSMK 0 . 200 0.200 0.200 0.2 00 NOCORR 0.950 0.950 0.950 0 .9 5 0 MONSMK 0.950 0 . 950 0.950 0.9 50 DAYSMK 0.950 0.950 0.950 0. 9 50 ADDSMK 0.950 0 . 950 0.950 0. 95 0 RHO PAP.AMETERS FOR LATE.' RESPONSE C A T E G O R Y 1 L N M N D N F N I O O O A O T O F N Y N £ C C D c I I i I G G 1 G NONSMK 0 . 800 0 .800 0 . 800 0 .8 0 0 NOCORR 0.0 50 0 .050 0 . 0 5 0 0. 050 MONSMK 0 . 0 5 0 0 . 0 5 0 0.050 0 . 0 5 0 DAYSMK 0. 050 0 . 0 5 0 0 . 0 5 0 0.0 50 ADDSMK 0. 050 0.050 0 . 050 0.0 50 " C O N T R O L " A T T I M Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission R E S P O N S E CATEGORY 2 L Y M Y ; D Y F Y T p O E A E T E C S N S Y S N S p C c D c I I I G G i G N O N S M K 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 N O C O R R C . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 M O N S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 D A Y S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 A D D S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 R H O P A R A M E T E R S F O R L A T E N T C L A S S " P R O G R A M " A T T I M E R E S P O N S E C A T E G O R Y 1 L N M N D N F N I O O O A O ; ^ F N Y N E C n D r I I I G O N O N S M K C . 3 0 0 C . 3 0 0 0 . 3 0 0 0 . 3 0 0 N O C O R R 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 C . 0 5 0 M O N S M K 0 . 0 5 0 0 . 0 5 0 C . 0 5 0 0 . 0 5 0 D A Y S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 A D D S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 R E S P O N S E C A ' T E G O R Y n L Y M Y D Y F Y I E O E A E ' -T’ E F s M S Y S N S E C C D • C T I ; I ' G ' C ; G N O N S M K 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 N O C O R R 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 M O N S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 D A Y S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 A D D S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 R H O P A R . A M E T E R S F O R L A T E N T C L A S S " P R O G R A M " A T T I M E Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. R E S P O N S E CATEGORY 1 ; L M M N ; D N F N i I O O O ' A O T O p N Y M E C r* D 1 C I : r I G G ; G N O N S M K 0 . 8 0 0 0 . 8 0 0 0 . 8 0 0 0 . 8 0 0 N O C O R R 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 M O N S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 D A Y S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 A D D S M K 0 . 0 5 0 0 . 0 5 0 C . 0 5 0 0 . 0 5 0 p . E S P o r I S E C A Y E G O R Y - L Y M Y D Y F Y I E O E A E T E F S N S Y S N S E C r^ D C I I I G G N O N S M K 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 N O C O R R 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 M O N S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 D A Y S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 A D D S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 R H O P A R A M E T E R S F O R L A T E N T C L A S S " P R O G R A M R E S P O N S E C A T E G O R Y T . N M N D N p M I O O O A O T 0 F N Y N p C C D C r I I sj G G N O N S M K 0 . 8 0 0 0 . 8 0 0 0 . 8 0 0 0 . 8 0 0 N O C O R R 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 M O N S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 D A Y S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . C 5 0 A D D S M K 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 0 . 0 5 0 A T T I M E Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 8 8 R E S P O N S E CATEGORY 2 L Y I M Y D Y F Y I E ! O E A E i T E F S N S Y S ' N S E C C D C r r I G G G N O N S M K 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 0 . 2 0 0 N O C O R R 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 M O N S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 D A Y S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 A D D S M K 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 0 . 9 5 0 G A M M A P A R A M E T E R S G A M M A S A R E U N C O N D I T I O N A L P R O B A B I L I T I E S O F M E M B E R S H I P I N E A C H L A T E N T C L A S S O F T H E S T A T I C L A T E N T V A R I A B L E C O N T R O L 0 . 5 0 0 P R O G R A M 0 . 5 0 0 D E L T A P A P A M E T E P . S D E L T A S A R E P R O B A B I L I T I E S O F L A T E N T S T A T U S M E M B E R S H I P C O N D I T I O N A L O N L A T E N T C L A S S D E L T A P A R A M E T E R S F O P L A T E N T C L A S S " C O N T R O L T I M E i N O N S M K 0 . - 7 0 0 N O C O R R 0 . 1 0 0 M O N S M K Q . 1 Q 0 D A Y S M K 0 . 0 7 5 A D D S M K 0 . 0 2 5 D E L T A P A R A M E T E R S F O R L A T E N T C L A S S " P R O G R A M " T I M E 1 N O N S M K 0 . 7 0 C N O C O R R 0 . 1 0 0 M O N S M K 0 . 1 0 0 D A Y S M K 0 . 0 7 5 A D D S M K 0 . 0 2 5 T A U P A R A M E T E R S T A U S A R E P R O B A B I L I T I E S O F L A T E N T S T A T U S M E M B E R S H I P A T T I M E T + l ( C O L U M N S ) C O N D I T I O N A L O N L A T E N T S T A T U S M E M B E R S H I P A T T I M E T ( R O W S ) A N D O N L A T E N T C L A S S M E M B E R S H I P T R A N S I T I O N P R O B A B I L I T I E S F O R L A T E N T C L A S S " C O N T R O L " Reproduced w ith perm ission of the copyright ow ner. Further reproduction prohibited w ithout perm ission. 1 8 9 RO W S R E P R E S E N T L A T E N T S T A T U S M E H B E P . S H I P A T T I M E 1 C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 1 N ! N M ! D A ! 0 o i o ; A d ; N c : n Y D S 0 ; s S s M p. M M M ; : k : r K K K N O N S M K 0 . 4 0 0 0 . 2 0 0 0 . 2 0 0 0 . 1 0 0 . 1 00 N O C O R R 0 . 0 0 0 0 . 5 0 0 0 . 2 0 0 0 . 2 0 0 0 . 1 0 0 MONSMK 0 . o c o 0 . 1 0 0 0 . 6 0 0 0 . 2 0 0 . 1 0 0 D A Y S M K 0.000 0 . 1 0 0 0 . 1 0 0 0 . 7 0 0 .1 0 0 A D D S M K 0.000 0 . 0 5 0 0 . 0 5 0 C . 10 0 . 3 0 0 RO W S R E P F E S E N T L A T E N T S T A T U S M E M 3 E R S H I ? A T C O L U M N S R E P R E S E N T LA T E N T S T A T U S l E M B E R S H I P N N M D A O : 0 O A D N c N Y n S 0 S S S M p M M M K p E E H. r T I M E 3 N O N S M K 0 . 4 0 0 0 . 2 0 0 0 . 2 0 0 0 . 1 0 0 0 . 1 0 0 N O C O R R 0 . 0 0 0 0 . 5 0 0 0 . 2 0 0 0 . 2 0 0 0 . 1 0 0 M O N SM K 0 . 0 0 0 0 . 1 0 0 0 . 6 0 0 0 . 2 0 0 0 . 100 D A Y S M K 0 . 0 0 0 0 . 1 0 0 0 . 1 0 0 0 . 7 0 0 0 . 1 0 0 A D D S M K 0 . 0 0 0 0 . 0 5 0 0 . 0 5 0 0 . 1 0 0 0 . 8 0 0 T R A N S I T I O N P P 0 5 A 5 I L I T I E S F OP LAI : e n t c l a s s RO W S R E P R E S E N ' T L A T E N :t S T A T U S M E M B E R S H I P A T C O L U M N S R E P R E S E N T LAi T E N T S T A T U S M E M B E R S H I P N N M D A O O O D : n C N I D S 0 S s S M p M M M K ; R K K K N O N S M K 0 . 4 0 0 1 0 . 2 0 0 0 . 2 0 0 0 . 1 0 0 0 . 10 0 N O C O R R 0 . 0 0 0 0 . 5 0 0 0 . 2 0 0 0 . 2 0 0 0 . 100 M O N S M K 0 . 0 0 0 0 . 1 0 0 0 . 6 0 0 0 . 2 0 0 0 . 1 00 D A Y S M K 0 . 0 0 0 0 . 1 0 0 0 . 1 0 0 0 . 7 0 0 0 . 100 A D D S M K 0 . 0 0 0 0 . 0 5 0 0 . 0 5 0 0 . 1 0 0 0 . 8 0 0 ' P R O G P A M Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 9 0 R O W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 2 C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 3 M N M D A ' O O O A D ! M C N Y D s : O S ! S S : M R M : M M K P. K K K N O N S M K 0 . 4 0 0 0 . 2 0 0 0 . 2 0 0 0 . 1 0 0 0 . 1 0 0 N O C O R R 0 . 0 0 0 0 . 5 0 0 0 . 2 0 0 0 . 2 0 0 0 . 1 0 0 M O N S M K 0 . 0 0 0 0 . 1 0 0 0 . 6 0 0 0 . 2 0 0 0 . 1 0 0 D A Y S M K 0 . 0 0 0 0 . 1 0 0 0 . 1 0 0 0 . 7 0 0 0 . 1 0 0 A D D S M K 0 . 0 0 0 0 . 0 5 0 0 . 0 5 0 0 . 1 0 0 0 . 8 0 0 I T E R A T I O N H I S T O R Y S T A R T I N G G - S Q U A R E D = 1 8 5 7 1 . 9 4 2 I T E R - M A D I T E R - M A T ) I T E R - M A D A T I O N A T I O N A T I O N 1 . 0 4 1 7 1 1 2 1 2 9 2 . 0 7 3 S 0 " ’ 7 2 8 1 3 . 0 4 5 3 4 4 5 5 5 3 4 . 0 2 0 2 8 1 9 4 3 4 c . 0 0 5 3 9 1 8 3 4 3 6 . 0 0 1 2 2 6 6 0 0 6 . 0 0 1 5 7 9 3 9 8 6 a . 0 Q 3 ' 7 8 6 4 ' ’ 3 ' 7 9 . 0 0 2 2 1 3 3 3 2 2 1 0 . 0 0 0 4 5 2 1 9 3 6 1 1 . 0 0 0 1 5 2 3 1 3 3 1 2 . 0 0 0 0 6 1 1 2 1 0 1 3 . 0 0 0 0 2 5 0 0 0 3 1 4 . 0 0 0 0 1 2 2 3 9 1 1 5 . 0 0 0 0 0 4 2 5 0 9 G - S Q U A P . E D = 2 4 9 . 2 6 6 W I T H 8 1 1 6 D E C R E E S O F F R “ W A R M I N G * * : B E S U R E T O I N T E P P R E T T H E L A T E N T C L A S S E S C A R E F U L L Y B A S E D O N T H E E S T I M A T E D P H O P A R A M E T E R S R E P O R T E D B E L O W . Y O U M A Y W I S H T O C H A N G E T H E L A B E L S Y O U P R E V I O U S L Y A S S I G N E D T O T H E L A T E N T C L A S S E S I N O R D E R T O M A K E T H E M C O N S I S T E N T W I T H Y C U R I N T E R P R E T A T I O N . L I T T L E R H O P A R A M E T E R S L I T T L E P . H O S A P E P R O B A B I L I T I E S O F R E S P O N S E S T O I T E M S M E A S U R I N G T H E S T A T I C L A T E N T V A R I A B L E C O N D I T I O N A L O N L A T E N T C L A S S M E M B E R S H I P R E S P O N S E C A T E G O R Y 1 V C .-I O P. N T 1 P. O L C O N T R O L 1 . 0 0 0 P R O G R A M 0 . 0 0 0 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 9 1 R E S P O N S E C A T E G O R Y 2 i v p ; ; A R : R O ; G ■ 1 P . i A m : C O N T R O L 0 . 0 0 0 P R O G R A M 1 . 0 0 0 * * W A R N I N G * * : E E S U R E T O I N T E R P R E T T H E L A T E N T S T A T U S E S C A R E F U L L Y B A S E D CM T H E E S T I M A T E D RHO P A R A M E T E R S R E P O R T E D B E L O W . Y OU MAY WISH T O C H A N G E T H E L A B E L S Y O U P R E V I O U S L Y A S S I G N E D T O T H E L A T E N T S T A T U S E S I N O R D E R T O M A K E T H E M C O N S I S T E N T W I T H Y OU R I N T E R P R E T A T I O N . B I G R H O P A R A M E T E R S B I G R H O S A R E P R O B A B I L I T I E S O F R E S P O N S E S T O I T E M S M E A S U R I N G T H E D Y N A M I C L A T E N T V A R I A B L E C O N D I T I O N A L O N L A T E N T S T A T U S , L A T E N T C L A S S , A N D T I ME R H O P A R A M E T E R S F O R L A T E N T C L A S S " C O N T R O L ” A T TIM E R E S P O N S E C A T E G O R Y 1 L M M N ' D N F N I O O O A O T O F M Y N E C C D C I I I G G G N O N S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 N O C O R R 0 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 M O N S M K 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 1 . 0 0 0 D A Y S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 A D D S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 R E S P O N S E C A T E G O R Y -1 L Y M Y D Y F Y I E O E A E T E F S M S ; Y s N S E C c D C I i I G G G N O N S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 N O C O R R 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 M O N S M K 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 D A Y S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 A D D S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . o c o Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. RHO PARAMETERS FOR LA TEN T CLASS "CONTROL " AT T I M E 2 R E S P O N S E CATEGORY 1 L M M N D N F N I O O C A O T O . F N Y N E r' c D C I T I G G j G i 1 NONSMK 1 . 0 0 0 1 . 0 0 0 1 . 000 ! 1 .00 0 NOCORR 0 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 000 MONSMK 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 1 . 000 DAYSMK 0 . 0 0 0 0 . 000 0 . 0 0 0 1 . 0 00 ADDSMK 0 . 0 0 0 0 . 0 0 0 0 . ooc 0 . 0 00 R E SPO N SE CA.'TEGORY n L Y M Y D Y F Y I E O E A E T E F S N S Y S N S E r 1 r 1 r n I I i r\ G NONSMK 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 00 NOCORR 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 .0 00 MONSMK 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 0 . o c o DAYSMK 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 ADDSMK 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 P.HO PARAM ETERS FOR LATENT CL A SS "CONTROL RE SPO N SE CATEGORY 1 u N M N D N : F N i . O O O A O T O F N ; 1* N E ! G c : D c I r I G G G NONSMK 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 NOCORR 0 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 MONSMK 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 1 . 0 0 0 DAYSMK 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 ADDSMK 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 AT TIME Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. R ESPONSE CATEGORY 2 L Y M Y D Y F Y I E O E ! A E T E F s N S i Y S i N S E C c D C I I I G G N O N S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 N O C O R R 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 M O N S M K 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 D A Y S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 A D D S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 R H O P A R A M E T E R S F O R L A T E N T C L A S S ’’ P R O G R A M R E S P O N S E C A T E G O R Y 1 L N M 11 D N zr v ! i o O 0 : A O T O 1 F N Y M ' E : c D C I : I I J G G N O N S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 N O C O P . P 0 . 0 0 0 1 . 0 0 c 1 . 0 0 0 1 . C 0 0 M O N S M K 0 . 0 0 0 C . 0 0 0 1 . 0 0 0 1 . 0 0 0 D A Y S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 A D D S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 R E S P O N S E C A T E G O R Y 2 L Y M Y D Y F Y I E O E A E T E F S N S Y S N S E c C D ; C I I i I G G : G N O N S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 N O C O R R 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 M O N S M K 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 D A Y S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 A D D S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 A T T I M E Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. P.HO PARAM ETERS FOR LATENT CL A SS "PROGRAM " AT T IM E 2 R E S P O N S E C A T E G O P ' N O N S M K N O C O R R M O N S M K D A Y S M K A D D S M K L N M N D N F M I O C O A O T O F : m Y : N c c C D C r I I G G G . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0.000 1.000 1.000 1.000 0.000 0.000 1.000 1.000 0.000 0.000 0.000 1.000 0.000 0.000 0.000 0.000 R E S P O N S E C A T E G O R Y L Y M Y D Y F Y I E 0 E .h E T F S N « L* S * I S E c D C T I I G N O N S M K . 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . C 0 0 N O C O R R 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 M O N S M K 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 D A Y S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 A D D S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 R H O P A R A M E T E R S F O P L A T E N T C L A S S " P R O G R A M R E S P O N S E C A T E G O P Y 1 L N m D N F M : 1 O 0 c A O T O F ; n ; Y N ; e : C 1 C D c I I I G G G N O N S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 N O C O R R 0 . 0 0 0 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 M O N S M K 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 . 0 0 0 D A Y S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 A D D S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 .T T I M E Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 9 5 R E S P O N S E C A T E G O R Y 2 L Y : m y ; d y ; F '{ I E O E A E T p F S N S Y S N S E C C D C I I I ; G r * VJ G j N O N S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 NOCOP.P. 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 M O N S M K 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 D A Y S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 0 0 0 . 0 0 0 A D D S M K 1 . 0 0 0 1 . 0 0 0 1 . 0 CO 1 . 0 0 0 G AM M A P A R A M E T E R S GAMMAS A R E ITNC M E M B E R S H I P I N E A C H L A T E N T C L A S S C O N T R O L 0 . 4 9 2 P R O G R A M 0 . 5 1 8 ■ O N D I T I O N A L P R O B A B I L I T I E S O F O F T H E S T A T I C L A T E N T V A R I A B L E D E L T A P A R A M E T E R S D E L T A S A R E P R O B A B I L I T I E S O F C O N D I T I O N A L O N L A T E N T C L A S S D E N T S T A T U S M E M B E R S H I P D E L T A P A R A M E T E R S FOP. L A T E N T C L A S S " C O N T R O L TI M E 1 T I M E 2 T I M E 3 N O N S M K N O C O R R M O N S M K D A Y S M K A D D S M K 0.871 0 . 1 0 9 0 . 008 0. 0 0 ^ 0 . 0 0 5 0 . 7 9 7 0 . 1 4 5 0 . 0 2 9 0 . 0 1 5 0 . 7 2 9 0 . 1 3 5 0 . 04 4 C . 024 0 . 0 1 3 D E L T A P A R A M E T E R S FOP. L A T E N T C L A S S " P P C G P . A M " TI M E 1 T I M E 2 T I M E 3 N O N S M K 0 . 3 9 3 0 . 3 4 9 0 . 7 7 9 N O C O P P . 0 . 0 ~ 8 0 . 1 1 4 0 . 1 5 0 M O N S M K 0 . 0 1 9 0 . 0 2 6 0 . 0 4 1 D A Y S M K 0 . C 0 4 0 . 0 0 “ 0 . 0 1 2 A D D S M K 0 . 0 0 1 0 . 0 0 4 C . 0 0 3 T A U P A R A M E T E R S T A U S A R E P R O B A B I L I T I E S O F L A T E N T S T A T U S M E M B E R S H I P A T T I M E T + l ( C O L U M N S ) C O N D I T I O N A L ON L A T E N T S T A T U S M E M B E R S H I P A T T I M E T ( R O W S ) A N D O N L A T E N T C L A S S M E M B E R S H I P T R A N S I T I O N P R O B A B I L I T I E S F O P . L A T E N T C L A S S " C O N T R O L " RO W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 1 C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T T I M E 2 N ! N M D 0 0 O A D N C M Y D S ! 0 S S S M p. M M M K R K K K Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 19G M O N S M K 0 . 9 1 5 0 . 0 5 7 0 . 0 1 9 0 . 0 0 5 0 . 0 0 3 N O C O R R 0 . 0 0 0 0 . 3 1 8 0 . 0 8 9 C . 0 4 7 0 . 0 4 6 M O N S M K 0 . 0 0 0 0 . 3 7 5 0 . 1 2 5 0 . 3 7 5 0 . 1 2 5 D A Y S M K 0 . 0 0 0 0 . 4 2 9 0 . 2 8 6 0 . 1 4 3 0 . 1 4 3 A D D S M K 0 . 0 0 0 0 . 0 0 0 0 . 0 0 0 0 . 2 0 0 0 . 8 0 0 R O W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P N N M D A O ; O O i A D ’ N C N ! Y D s O S S S M R M M M E P. K K N O N S M K 0 . 9 1 4 0 . 0 5 6 0 . 0 1 8 0 . 0 0 5 0 . 0 0 6 N O C O R R 0 . 0 0 0 0 . " 9 1 0 . 1 2 4 0 . 0 5 1 0 . 0 3 5 M O N S M K C . 0 0 0 0 . 6 2 2 0 . 2 1 7 0 . 1 0 9 0 . 0 5 2 D A Y S M K 0 . 0 0 0 0 . 2 0 8 0 . 2 0 6 0 . 4 8 1 0 . 1 0 4 A D D S M K 0 . 0 0 0 0 . 3 2 1 0 . 1 6 1 0 . 1 6 1 0 . 3 5 7 T R A N S I T I O N P R O B A B I L I T I E S F O R L A T E N T C L A S S R O W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T C O L U M N S R E P R E S E N T L A T E N T S T A T U S »: »? m M E M 3 E P S H I P 0 ^ ^ f . r t D N C N Y D 5 O S S S M P. M M M K R K K K N O N S M K 0 . 9 4 6 0 . 0 4 1 0 . 0 1 1 0 . 0 0 1 G . 0 0 1 N O C O R R 0 . 0 0 0 0 . 9 2 7 0 . 1 2 3 0 . 0 3 9 0 . 0 1 2 M O N S M K 0 . 0 0 0 0 . 6 5 1 0 . 2 0 3 0 . 0 9 9 0 . 0 4 9 D A Y S M K 0 . 0 0 0 0 . 0 0 0 0 . 5 1 0 0 . 2 4 5 0 . 2 4 5 A D D S M K 0 . 0 0 0 0 . 0 0 0 1 . 0 0 0 0 . 0 0 0 0 . 0 0 0 R O W S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P A T C O L U M N S R E P R E S E N T L A T E N T S T A T U S M E M B E R S H I P N N M D A G O O A D N r* * r • * D S O S S S : m P. M M M : K p. k ; k K N O N S M K 0 . 9 1 7 0 . 0 5 9 0 . 0 1 6 0 . 0 0 5 C . 0 0 3 N O C O R R 0 . 0 0 0 0 . 8 4 8 0 . 1 2 7 0 . 0 1 7 0 . 0 0 8 M O N S M K 0 . 0 0 0 0 . 3 3 9 0 . 4 3 2 0 . 1 4 4 0 . 0 3 6 D A Y S M K 0 . 0 0 0 0 . 3 0 5 0 . 1 3 9 0 . 1 3 9 0 . 4 1 7 A D D S M K 0 . 0 0 0 0 . 5 0 0 0 . 0 0 0 0 . 2 5 0 0 . 2 5 0 ★ ★ ★ ★ ★ ★ P R O G R A M F I N I S H E D : T h u O c t 2 5 2 0 : 2 2 : 3 6 2 0 0 1 T i m e : 6 3 3 T I M E 2 A T T I M E 3 P R O G R A M " T I M E 2 A T T I M E 3 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. APPENDIX D 1 9 7 SAS CODES •Censored normal quadratic model - unconditional model; c r o c o ra ; daoa-cc-r: cat p 1 oc =s op c a to ta t^ s c s c at= scf; var smkstac* scikstgtZ ST.kstgti. •ir.dep TI T3; rr.dep G1-G3; model c r o r m ■ •Censored normal quadratic model - conditional model; p ro c t r a ] data=smk o u t p l o t = s r c p c u ts ta t = s r c s :u : = s r : f . id id; var sm k stg tl smkst go 2 smkstgt 3 ; ' m d e p 71 - 73 ; m d e p Z 1 - 12 , max 3 ; r.groups 3 ; ord er 2 2 2; r i s k sex c la c k l a t i n asiar. o th er prog SE3 3?AB 2PAZ ir.tsmk i s t a r t I . . . - * --------- ' *51 j 3 4 c 3 4 j 33 * 3 33 2 4 . 2 “ 3 1 1" 9 *3 <7 RAJ PL-2 7 d? 2 ?, 2?... J . 'Later.*; Zrcwth Ti a ; e : t o r i e s : : Z m o k m g '. 'Z e r . s r r e i L irm a l Vodel * Non-linear Mixed Model - logistic; p ro c r.lmixed data=smk, parms beta2=3 o2a -2, e t a = betaO • u; e x p e ta - e x p ' e t a ' ; p = e x p e t a / . i- e x p e ta : ; model x - bir.cmial 1. p random u - normal 3,s2u.' s u b ] e c t= s c h o o l; p r e d i c t e t a o u t= e ta; • e s t i m a t e ' L / b e t a l ' l / b e t a l ; run; * GEE model - independent; p r o c genmcd data=mult robs descending; c l a s s id sch o o l; model sm.kadd = t imec g en d er white groupc ir .t r l psm.ktl fsm.ktll pfsmktl t im.ec *t im.ec groupc*timec g r : u p c r t imec*t imec d = b m type! w a id w a ld c i; re p ea te d sub ; ecr - id sc h o o l tyce=i:id, t i t l e 1 3E~ l o g i s t i c model, with, i n t e r a c t i o n t e r m s ' , run,- Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 9 8 APPENDIX E M p l u s s c r i p t s (a) Test for mediation TITLE: DiiiSert.i: ::r. Research. TCP? Panel Lata n = Z. 12 1 Mediator Model - P a r a l l e l process ? roaran e f f e c t s mediated through growth of in te n t Cutccm.e v a r i a b l e is l i f e t i m e c i g a r e t t e use. M u ltile v e l Growth Curve Model - Mediator Model Level 1: Growth Curve of L ifetim e smoking Growth Curve of i n te n t io n Level 2: Program on growth f a c t o r s of in te n tio n LATA- FILE IS 'o- Chaoyar.gLi 1 chaoyang data tops smki.nt r OF MAT .s 1 ■ ? F1 . IF2, IrU.Z, VARIABLE Vanes a re d u l l d u l ! d u l l tlqZC tZgZ-i d e l ! ::;te:;t ; intent: intent: GPCL'P GENDER race white l a t i n as oar. other school FPEEMEAL; C l u s t e r i s s c h o o l ; Between is group. m i s s :::i is VSEVA? = in te n t 1 i n t e n t ! scho ol group lEFINE : : e c n .< 1 = 1 1 cZ 1. : vsnkZ = t Z d 2 . : esmk ! = t 3 q ll , ANA LI 3 I > TV PL - M FA.'3 T P . V CT V?. E t w; I 7EP.A7 1 CVS = Z d d ; 4 betv;een * l i b 3V i n te n t ; i n t e n t ! 1. s i b BY in te n t 1 • C.5 i n t e n d ' i n t e n t ; • 1 iZb BY l i t e s n k l 1 i f e s n . k ! •1; sZb BY 1 if e sm kI ■ 0 .5 li f e s m k Z - 0 l i f e s m k ! [ i n t e n t 1 - i n t e n t 3-0 11 fes mkl - 11 fesmk! O ’ ’l i b s i b iZb sZb g r o u p ; , i l b - s i b CM group; iZb sZb C . \ ’ group,- •w i t h i n * i l w BY i n t e n t 1 i n t e n t 2•1; s l w BY i n t e n d : 0 . 5 i n t e n t 2 ■ * 0 m t e n t 3 - 1 . 0 iZw BY 11 fes mk 1 - 1 1fe s m k 3 d ; sZw BY 1 i f esmkl • 0 . 5 IifesmkZ^O l i f e s m k ! = sZw Oil 1 1W slw; lZ w ON 1 1w slw; CVTPVT : STANDARD I ZED ; dat • 1 . 0 ; Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 1 9 9 (b) . Test for interaction TITLE: D i s s e r t a t i o n Research. TOFP Panel Tata . rt=D . 0 3 31 Moderator Model F n ends' Smoking vs. R efusal Se1f -E ffic ac v Cutctme v a r i a b l e is l i f e t i m e c i g a r e t t e use. DATA. FUR IS ’c: Cha tyar.gL 11 chaoyang data t :c p f sm.ksec . dan f:? m a t u ?F5 :. ~ fi, i f i if-i : VARIABLE: Dames a re 1 f sm.kl - 1 f smk3 F3MFC1 - FSMEC3 RSEC1-RSEC3 CROC? GEDDER race white l a t i n as ran other school FREEMEAL; MISS IDG IS . ; 7 3 E 7 A E = FSM r'Cl FS M r'Cl RS EC1 RSEC3 ger.der white group f s e l f s e l life sm k l lifesm k!. DEFIDE. f s e l =fsmkCl•rse Cl f s e != f smkC!* rse 3! f se 3 = fsmkCl•rseC3 l i f esmk 1 = 1 f smk 1 11fesmk! = 1fsmk!; 11 f esmk3 = 1 f sm.k.3 ; AliALi^Iw- TYPE = Mz.AD3 . RL'!7L’R =. missir.u HI. i ce?a :i c d s=:~ :: :. MODEL: l e v e l l BY fsmkCl fsmkC3-l, t r e n d l BY fsmkCl 1.5 fsmkCl 3 fsmk!3*l l e v e l ! BY r s e c l r s e c 2 ■ 1 ; t r e n d ! BY r s e c l - 0.5 r o e c!-2 r s e c l -1 0; l e v e l ! BY f s e l f s e i : l; t r e n d 3 BY f s e l — 3.5 f s e S ■1 f s e l *1.3; level-; BY lif~ sm k l lifesm k! 1; trend-; BY lif e s m k l- 1.5 1 i f esmk! • 0 1 : f esc..-: • - 1 ifsT.hr I f so.-: r 2 2 r s e s l r s e r ! :3 f s e l fse l -I life sm k l 1 i f esc-; 3 - C i , i l e v e l l t r e r .il l e v e l ! t r e n d l le**el3 tren d s level-; t r e n d ; ] . ; C E D ’DER white g: sup! , trend-; C D * t r e n d l tr e n d ! tren d -, trendd 3D gender w hite group, trer.u-i 3D * l e v e l l l e v e l ! lev e l-, level-; CD t r e n d l tr e n d ! tren d !, level-; CD gender white group level-; CD levtrll l e v e l ! le v e l!. (C) Latent growth mixture modeling TITLE: TOFP ;n =I,C 5 3 '. Growth Mixture Modeling fo r Sacking S ta g e s Smoking Stag es, o r d i n a l s c a le s : 1 =r.on - smokers : ! =r.on - cur re n t smokers. 3= current smokers m o n th ly - . •;=regular smokers - d a ily m 5 ^a d d ic te d smokers. Time was c e n te re d a t time 2. Q u a d ratic model u n c o n d it i o n a l model. DATA: FILE IS 1 c : cha oya.ng - cha o 1 mplus tope s m k s tg .d a t'; FORMAT IS f d * 1 C f 1 . f A 1 . Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission. 2 0 0 Reproduced w ith perm ission of the copyright ow ner. F urther reproduction prohibited w ithout perm ission.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Emotional intelligence and smoking risk factors in early adolescents
PDF
Correlates of smoking behaviour among Egyptian adolescents
PDF
Adolescent exposure to alcohol advertising: A prospective extension of Strickland's model
PDF
Implicit cognition and dissociative experiences as predictors of adolescent substance use
PDF
Cigarette smoking, metabolic syndrome, and subclinical atherosclerosis
PDF
Understanding visceral fat, subcutaneous abdominal fat, and total body fat in children: Longitudinal changes and relations to metabolic risks of type 2 diabetes
PDF
Aggression and victimization as predictors of drug use in early adolescence
PDF
Adoption and implementation of evidence-based tobacco use prevention curricula and programming
PDF
The relationship between high school anti-smoking policies and adolescent cigarette use
PDF
Stressful life events, social support, and smoking behaviors in United States And Chinese adolescents: A cross-cultural comparison
PDF
Contributions of interpersonal communication within a mass media environment: Exploring the role of social influence in a statewide tobacco control media campaign
PDF
Peer influences, depression and smoking
PDF
Multiple methods of measuring peer influence and peer selection for adolescent smoking
PDF
Correlates of compliance with mammography screening guidelines among low-income Latinas: An exploratory study
PDF
The influence of future orientation on the relationship between condom use and violence among homeless/runaway youth: Implications for the prevention of HIV and violence
PDF
Social network and group self-identification predictors of school violence
PDF
Tobacco marketing and smoking behavior among adolescent girls across ethnicity
PDF
Antioxidants and risk of myocardial infarction and cancer in a cohort of middle-aged finnish men: the kuopio ischemic heart disease risk factor study
PDF
Self-help and problem-proneness as predictors of weapon carrying among adolescents
PDF
Predictors of adolescent depressive symptomatology
Asset Metadata
Creator
Li, Chaoyang
(author)
Core Title
Patterns and risk factors for adolescent smoking progression
Degree
Doctor of Philosophy
Degree Program
Preventive Medicine - Health Behavior Research
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
health sciences, public health,OAI-PMH Harvest,psychology, psychometrics
Language
English
Contributor
Digitized by ProQuest
(provenance)
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c16-201642
Unique identifier
UC11339023
Identifier
3065814.pdf (filename),usctheses-c16-201642 (legacy record id)
Legacy Identifier
3065814.pdf
Dmrecord
201642
Document Type
Dissertation
Rights
Li, Chaoyang
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
Tags
health sciences, public health
psychology, psychometrics