Close
About
FAQ
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
/
The dynamic relationship of emerging adulthood and substance use
(USC Thesis Other)
The dynamic relationship of emerging adulthood and substance use
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
THE DYNAMIC RELATIONSHIP OF EMERGING ADULTHOOD AND
SUBSTANCE USE
by
Nadra Erin Lisha
A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PREVENTIVE MEDICINE: HEALTH BEHAVIOR)
May 2012
Copyright 2012 Nadra Erin Lisha
ii
EPIGRAPH
"C'est l'histoire d'un mec qui tombe d'un immeuble de cinquante étages, et qui au fur et à
mesure de sa chute se répète sans cesse pour se rassurer : Jusqu'ici, tout va bien…
Jusqu'ici, tout va bien… Mais l'important c'est pas la chute… C'est l'atterrissage."
- La Haine
iii
DEDICATION
To my grandmothers, Dorothy Gould (1917 - 1999) and Marie Lisha (1923 - 2011).
iv
ACKNOWLEDGEMENTS
First and foremost, I would like to thank my committee chair, Dr. Steve Sussman,
for your help and guidance. I feel very fortunate to have had such a wonderful and
supportive advisor. Thank you for your great sense of humor (even if I am not always
sure when you are joking). Thank you for always being there to talk to me when I
needed you, whether it was school-related or life-related.
I would also like to thank the other members of my dissertation committee, Dr.
Adam Leventhal, Dr. Ping Sun, Dr. Louise Rohrbach and Dr. Richard John, for their time
and assistance. Dr. Leventhal (I guess I’m’ supposed to call you Adam now…), thank
you for always giving me advice (even when I did not want any) and for truly caring
about my future.
Thank you to my parents for your wisdom and encouragement. I cannot express
how thankful I am to have you as my parents. We will always stick together through the
good times and the bad, and no matter what I know you will make sure I keep my head
up. I just hope I can live up to the person you have taught me to be. You widened my
scope of the world through travel, education, and love, and I will forever be thankful for
that.
Thank you to Doug, for all your love and support. We have been through a lot
over the years, and I cannot imagine them without you. You challenged me to live in a
new way, to think in a new way, and though I have been resistant at times, I thank you
and continue to learn every day. Perhaps one day I will know how to relax properly.
Now time for a trip!
v
Thank you Marny Barovich, without you I probably would have taken all the
wrong classes and would still be chasing my tail in circles. To Ryan Wilkerson - without
you and my trusty red bag the office would be so much less fun.
Thank you to all my great friends for your encouragement, and mostly for being
so entertaining. It makes me happy to know that I can always find someone to hike with
when I need an escape.
vi
TABLE OF CONTENTS
Epigraph ii
Dedication iii
Acknowledgements iv
List of Tables viii
List of Figures ix
Abstract xi
Chapter 1: Introduction
Specific Aims 1
Background and Significance 3
Introduction to Dissertation Studies 28
Chapter 2: Study 1: Psychometrics of the IDEA
Chapter 2 Abstract 30
Introduction 30
Present Study 31
Method 32
Participants and Procedures 32
Survey Measures 33
Statistical Analyses 36
Results 37
Discussion 44
Chapter 3: Study 2: The relationship of emerging adulthood
trajectories to drug use, and other behavioral outcomes across
three waves of data
Chapter 3 Abstract 52
Introduction 53
Present Study 60
vii
Method 64
Participants and Procedures 64
Survey Measures 64
Statistical Analyses 68
Results 79
Discussion 103
Chapter 4: Discussion 114
Limitations 116
Implications for Future Research 117
Bibliography 119
Appendix 134
viii
LIST OF TABLES
Table 1. Full IDEA items and dimensions. 21
Table 2. PCA factor loadings after orthogonal
rotation.
40
Table 3. Correlations between factors. 41
Table 4. Regressions of the IDEA full scale and
subscales with other variables.
43
Table 5. Reduced emerging adulthood measure. 65
Table 6. Descriptive statistics of IDEA items across
time.
80
Table 7. Correlations matrix among items across
time.
81
Table 8. Model fit of various invariance models. 82
Table 9. Model fit criteria for 1 to 7 class models in
LCGA without covariates.
90
Table 10. Model fit criteria for 1 to 7 class models in
LCGA without covariates (program, age gender).
90
Table 11. Association between latent class
membership and baseline demographics.
97
Table 12. Class membership (N = 1,682) and 100
endorsement frequencies of the Time 3 correlates
using the best fitting LCGM solution.
Table 13. Differences for Time 3 outcomes by latent
growth class.
102
ix
LIST OF FIGURES
Figure 1. Four main substance use trajectories. 7
Figure 2. Median age at first marriage (NPR). 10
Figure 3. College Enrollment Among 18-24 Year Olds. (Fry,
2009) 11
Figure 4. IDEA factor structure model per Reifman, Arnett &
Colwell (2007). 38
Figure 5. Basic two-factor, three time point latent growth model. 70
Figure 6. Graphical representation of the associative growth
mixture model. 74
Figure 7. Graphical representation of the associative growth
mixture model with covariates. 76
Figure 8. Observed latent mean trajectories without covariates
(N = 100). 85
Figure 9. Estimated mean trajectories without covariates (N =
100). 86
Figure 10. Overall sample and estimated mean. 87
Figure 11. Observed mean trajectories with covariates (N = 100). 88
Figure 12. Estimated mean trajectories with covariates (N = 100). 88
Figure 13. Final multiple group model of emerging adulthood
trajectories over time adjusting for covariates.
91
Figure 14. Observed individual values with estimated latent class
means for Class 1. 92
Figure 15. Fitted individual values with estimated latent class
means for Class 1.
92
Figure 16. Observed individual values with estimated 93
x
Figure 17. Fitted individual values with estimated
latent class means for Class 2.
93
Figure 18. Observed individual values with estimated latent
class means for Class 3.
94
Figure 17. Fitted individual values with estimated latent class
means for Class 2.
94
xi
ABSTRACT
The present dissertation project examined emerging adulthood in the context of
substance use behavior. It is presently recognized that individuals pass through a
transitional period between adolescence and young adulthood - “emerging adulthood.”
Continuation high school youth have demonstrated to show early entry into emerging
adulthood-type developmental tasks. The first study examined the psychometric
properties of the Inventory of Dimensions of Emerging Adulthood (IDEA) in a
population of continuation high school students of Southern California. The IDEA was
developed in an attempt to capture the psychosocial attributes of this unique
developmental period. A 21-item version of the IDEA was evaluated, and it was
determined that the measure was composed of three dimensions and demonstrated high
internal consistency. In addition, construct validity was assessed and indicated that the
measure, as expected, was correlated with demographic characteristics, risk behaviors,
and other psychological measures. It was concluded that the IDEA is a practical
instrument for measuring emerging adulthood in at-risk populations. Not every
individual experiences emerging in the same fashion; emerging adulthood is
characterized by its great heterogeneity. As such, the second study used a shortened 8-
item IDEA measure to examine trajectories of emerging adulthood. The items were
examined for factorial invariance across three time points of data collection. A shortened
5-item measure was found to be factorially invariant and was used in subsequent analysis.
A latent growth curve model analysis indicated a sample that was growing in emerging
adulthood status over time. Next, a latent class growth analysis was used to identify
xii
distinct trajectories. Three such trajectories were extracted from the data. These latent
classes were then examined in terms of baseline characteristics as well as Time 3
behavioral outcomes. Large differences were found at baseline, but appeared to dissipate
over the two-year period. These findings are expected to provide valuable implications
for drug prevention programs directed at at-risk populations. It appears that this
population is entering emerging adulthood earlier than other normative youth. As such, it
might be beneficial to target these students earlier for risk behavior prevention efforts.
Future research should examine emerging adulthood over a longer period of time, over
additional developmental periods, and among diverse populations. If over longer periods
of time differential outcomes are exhibited, it might be possible to distinguish these
groups based on baseline characteristics, and perhaps to target those with the initial most
risky trajectory.
1
CHAPTER 1: INTRODUCTION
Specific Aims
Although a substantial amount of research has examined the etiology of drug use
in the United States, few studies have examined the relationship between emerging
adulthood and drug use etiologies. The 21-item Likert-type scale of emerging adulthood,
the Inventory of the Dimensions of Emerging Adulthood (IDEA; reduced form), intends
to tap the themes associated with the transition to adult status. In Study I, for the first
time to our knowledge, the psychometric properties of this scale are investigated in
continuation high school youth. This study provides valuable information for use of the
IDEA measure in a continuation high school population. The correlation of the measure
and its subscales with demographics, risk behaviors, and psychological measures were
examined. A reduced 8-item measure will be used subsequently to predict trajectories of
emerging adulthood longitudinally in Study II. Class membership for emerging
adulthood trajectory was then be regressed on the measures from Study I and some
additional measures. Findings from Study II are expected to contribute to the current
youth developmental literature on emerging adulthood by elucidating the dynamic
linkage between emerging adulthood trajectories and various behavioral correlates. The
specific aims for the dissertation project include:
Dissertation Study I
1. Assess the psychometric properties of the Inventory of the Dimensions of
Emerging Adulthood (IDEA) in a sample of continuation high school students
using exploratory factor analysis (EFA) procedures
2
a. Investigate differences in how the emerging adulthood period is
experienced by a non-white at-risk population;
b. Explore the reliability of the emerging adulthood construct in a non-white
at-risk population;
c. Examine the factor structure of the IDEA, and
d. Explore the predictive validity of the IDEA by examining relations of each
factor with variables that might indicate emerging adulthood transitions,
such as demographic characteristics, risk behaviors, and indicators of
psychological maturation.
Dissertation Study II
2. Assess the predictive utility of emerging adulthood trajectories on future
demographic and risk behavior correlates in continuation high school students
a. Examine the factorial invariance of the 8-item emerging adulthood
measure
b. Examine if individuals can be classified into discrete latent groups based
on the emerging adulthood status
c. Regress both latent class membership on measures from Study I
(demographics, and risk behaviors)
d. Compare regression results from exploratory (latent classes) and pre-
determined classes
3
Background and Significance
Adolescent Drug Use as a Public Health Concern in the United States
Adolescent substance use remains a significant health risk in the United States.
According to the most recent Monitoring the Future (MTF) national survey on drug use
in secondary school students, an illicit drug (i.e., marijuana and hard drugs) epidemic
persists (Johnston, Bachman, & Schulenberg, 2009). While in the early 1990s use rates
were higher in the college-age group than they were among secondary school students, by
the late 1990s, the highest rates of use were found in the late secondary school years
(Johnston et al., 2009). Also, in 2009, the rank order for ages of annual prevalence of
using any illicit drug was: 12
th
graders (37%), college students (36%), 19- to 28-year-olds
(33%), 10th graders (29%), and 8th graders (15%). In 2009, among 12
th
graders, 44.1%
of students will have at least tried some type of illicit drug, 46.7% will have at least tried
marijuana, 54.0% will have drunk some alcohol, and 28.3% will have smoked at least one
cigarette.
Among adolescents, the leading causes of death have remained constant
throughout the period 1999-2006: accidents (unintentional injuries) (48 %), homicide (13
%), suicide (11%). Motor vehicle accidents accounted for almost three quarters (73 %)
of all deaths from unintentional injury (Miniño, 2010). In 1996, young people, aged 16
to 24 were involved in 28 percent of all alcohol-related driving accidents, while they
make up merely 14% of the U.S. population (Campbell, Zobeck, & Bertolucci, 1996).
Thus, adolescents are both over-represented in driving accidents involving alcohol and in
4
drinking driver injuries and deaths (NIAAA, 1997). The increasing numbers of youth
homicides are among the social and criminal justice problems often linked to adolescent
substance abuse (Justice, 1996), as are suicides among adolescents (Moscicki, 1995).
Thus, disproportionate numbers of youth involved with alcohol and other drugs face an
increased risk of death through accident, homicide, and suicide.
Trajectories of Alcohol and Drug Use from Adolescence to Adulthood
Not all individuals who try a drug will become regular users, just as not all adult
non-users will have never tried drugs. It is only through careful examination of the
delicate interplay between physiological reward systems, affective and cognitive
individual differences, and social and environmental influences that we will predict drug
use vulnerability and sustainability. Thus, an unmistakable need to continue the
exploration of drug use predictors exists. This new research should be done among
diverse populations to better inform prevention science and thus enhance the overall
wellbeing of our present society.
Developmental and psychological perspectives suggest the critical importance of
identifying antecedents of drug and alcohol use (Hawkins, Catalano, & Miller, 1992; D.
B. Kandel, Kessler, & Margulies, 1978). Emerging adulthood developmental trajectories
have not been studied in terms of future behavioral and psychological outcomes. This is
of particular significance in at-risk youth as they are more likely to engage in alcohol and
drug use behaviors. Thus, the current sample of continuation high school students is
highly valued.
5
Evidence demonstrates that substance use and substance use disorders follow
systematic trends and age-related patterns. Typically onset occurs during adolescence,
peaks in use and diagnosed disorders during emerging adulthood (Arnett, 2000a), and
then subsequently declines in use post-mid-twenties (Bachman, Wadsworth, Johnston, &
Schulenberg, 1997; Chen & Kandel, 1995; Sher, Grekin, & Williams, 2005). While
overall trends are visible, there is still substantial heterogeneity in the initiation, and
development of alcohol and drug use and substance use disorders. This heterogeneity
may be indicative of etiological correlates; there are a variety of different correlates that
might reflect different types of trajectories (Chassin, Flora, & King, 2004). Some
research indicates that alcohol use disorders may differ in their antecedents in relation to
the course (“developmentally limited” emerging adulthood vs. persistent) (Zucker,
Fitzgerald, & Moses, 1995), stable traits, and the age of onset (Cloninger, 1987). While
various models for studying alcohol use involvement over the course of the lifespan have
been proposed, developmentalists generally support a typological, trajectory-based
approach (Sher et al., 2005). In an effort to understand the etiological variables that are
responsible for substance use trajectories, it might be important to distinguish differing
groups of individuals and how they follow different trajectories over time in comparison
to those who do not engage in substance use at all over time.
Thus, researchers have attempted to identify groups who follow particular
developmental courses of substance use, and compare them with other groups of
individuals in terms of the various antecedents and consequences that are related to the
particular substance use trajectory (Bennett, McCrady, Johnson, & Pandina, 1999;
6
Chassin, Pitts, & Prost, 2002; D. B. Kandel & Chen, 2000; J. Schulenberg, O'Malley,
Bachman, Wadsworth, & Johnston, 1996; J. Schulenberg, Wadsworth, O'Malley,
Bachman, & Johnston, 1996; J. S. Tucker, Orlando, & Ellickson, 2003). See Figure 1.
The normative non-user/stable low-user course begins at low levels of use and increases
moderately over time. This group is characterized by low levels of delinquency,
depressive symptoms and negative affect at older ages. The early onset, chronic or high-
user (early heavy) course includes individuals who begin using substances at a young
age, peak through emerging adulthood, and then decrease. This group is characterized by
the highest levels of delinquency, depressive symptoms and negative affect later in their
developmental course. The late-users course is characterized by slow early onset of use
followed by slightly lower than normative levels of use. This course is characterized by
low levels of delinquency, depressive symptoms, and negative affect. Lastly, individuals
in the later onset course exhibit extremely low to nonexistent levels of use followed by a
steady moderate increase in use until use is high. These trajectories are defined as
follows (Marti, Stice, & Springer, 2010): 1) normative non-user/stable low-user course,
2) early onset, chronic or high-user (early heavy) course, 3) late users course, and 4)
later-onset heavy user course. These individuals are characterized by moderate levels of
delinquency, depressive symptoms and negative affect.
7
Figure 1. Four main substance use trajectories (Marti et al., 2010).
Some associations with developmental milestones have been correlated with more
specific drug or alcohol use. For example, Bachman et al. (1997) found that moving out
of the parents’ home and becoming independent from adult supervision was positively
associated with increased alcohol use, while getting married and having children was
associated with declines in drinking. In addition, early age of onset of substance use has
been found to be associated with higher likelihood of substance use problems in
adulthood (Hawkins et al., 1992; Hawkins et al., 1997; D. Kandel, Simcha-Fagan, &
Davies, 1986). Frequent relocation of residence has been found to be associated with
earlier age of substance use for marijuana, hallucinogens and cocaine use (DeWit, 1998).
Other factors associated with development have been found to be associated with
substances use; such as parental educational attainment, family structure and having a
family member who is a substance user, have all been found to be associated with earlier
8
initiation into substance use for youth (Blackson et al., 1999; Ellickson, Tucker, & Klein,
2001; Rohde, Lewinsohn, Brown, Gau, & Kahler, 2003).
Various risk factors have been identified as being associated with substance use
trajectories. One example of a risk factor associated with both early onset and
persistence of cigarette smoking is family history of substance use (Chassin, Presson,
Pitts, & Sherman, 2000). Other studies have found that parental alcoholism is associated
with early onset of drinking behaviors (Chassin et al., 2002; Dawson, 2000), as well as
with persistent alcohol use disorders (Jackson, Sher, & Wood, 2000). Intrapersonal
characteristics such as sensation seeking and impulsivity have also been found to be
correlated with alcohol use trajectories (Bennett et al., 1999; Chassin et al., 2002). It is
possible that developmental trajectories might also be associated with alcohol and
substance use. For this reason, exploring how individuals experience and pass from
adolescence to adulthood is important. The investigation of developmental trajectories
might give us some indication of how substance use develops over time.
Emerging Adulthood Concepts and Measure
In most industrial societies just forty years ago the average 22 or 23 year-old
individual was married, had at least one child and was buying a house (Arnett & Tanner,
2006). Of course, today things are different. In industrial societies, the period from the
late teens until the mid-20s is now a time where both young men and women explore,
experience instability, and take time to reach full adulthood (Arnett & Tanner, 2006).
Young people are now expected to rely on their own resources and sense of agency (i.e.
9
subjective awareness of performing, and controlling one's own volitional actions);
individualization during the time between adolescence and adulthood has increased as
institutionalization has decreased (Côté, 2000; Heckhausen, 1999; Mayer, 2004). The
road to adulthood in the 21
st
century means something different to young people than it
did previously. Today, young people must demonstrate self-sufficiency, and the
capability to make choices independently while simultaneously considering a multitude
of possibilities (Arnett, 1998).
Arnett (2000a) developed the theory of emerging adulthood in an attempt to
conceptualize the development of young people in today’s industrialized societies. This
period encompasses approximately ages 18 to 25, and recognizes this time as a distinct
period of development. It is the unique time between adolescence and adulthood.
Demographics of the emerging adulthood period. The last forty years has
presented the United States and other industrialized nations with tremendous changes in
the demographic characteristics of emerging adults. Changes in age of marriage, timing
of parenthood, and participation in higher education have seen some of the largest
differences. Median age of first marriage rose steeply between the 1960s and the mid-
1990s, where it began to level off (NPR). While men and women have changed in a
parallel fashion, men continue to typically marry approximately two years later than
women. In 1960 the median marriage age was 22.8 for men, and 20.3 for women, while
in 2005 it rose substantially to 27.1 and 25.3 for men and women respectively. See
Figure 2.
10
Figure 2. Median age at first marriage (NPR).
The age at which individuals have their first child has also risen. In Europe,
mothers are currently on average 26 to 29 at the first childbirth, compared to 23 to 25 in
the 1970s (Schoonbroodt & Jones, 2007). This same trend can be found in Asia, Japan,
and the United States, and seems to be spreading to developing countries such as China,
and Turkey.
In addition, rates of participation in higher education have increased in the United
States. Part of the reason individuals are more likely to marry and have children later
might be due to increased participation in higher education. Over many decades college
enrollment has been rising at both two- and four-year colleges. In 2008, a record number
of 18 to 24-year olds were enrolled in college
nation’s youth in that age group. See Figure
not relegated to the United States; a similar pattern can be found in other industrialized
nations (Arnett & Tanner, 2006)
Figure 3. College Enrollment Among 18
Lastly, one of the most remarkable characteristics of the demographic outline of
emerging adulthood is their extremely high frequency of changing residence. Between
the ages of 20 and 24, right in the middle of emerging adulthood, is the peak rate of
residential change (Arnett & Tanner, 2006)
adults has evolved over the last half century. The period after high school, roughly from
18 through the mid-20s has changed; it is no longer a period of stability and taking on
adult roles (e.g. marriage, parenth
exploration, and feeling in
year olds were enrolled in college (Fry, 2009); almost 40 percent of the
nation’s youth in that age group. See Figure 3. This pattern of increasing enrollment is
not relegated to the United States; a similar pattern can be found in other industrialized
Tanner, 2006).
. College Enrollment Among 18-24 Year Olds (Fry, 2009).
Lastly, one of the most remarkable characteristics of the demographic outline of
hood is their extremely high frequency of changing residence. Between
the ages of 20 and 24, right in the middle of emerging adulthood, is the peak rate of
(Arnett & Tanner, 2006). Thus, the demographic picture of emerging
adults has evolved over the last half century. The period after high school, roughly from
20s has changed; it is no longer a period of stability and taking on
adult roles (e.g. marriage, parenthood, a career, and a mortgage), but one of instability,
exploration, and feeling in-between.
11
most 40 percent of the
. This pattern of increasing enrollment is
not relegated to the United States; a similar pattern can be found in other industrialized
Lastly, one of the most remarkable characteristics of the demographic outline of
hood is their extremely high frequency of changing residence. Between
the ages of 20 and 24, right in the middle of emerging adulthood, is the peak rate of
Thus, the demographic picture of emerging
adults has evolved over the last half century. The period after high school, roughly from
20s has changed; it is no longer a period of stability and taking on
ood, a career, and a mortgage), but one of instability,
12
The five domains of emerging adulthood.
Five primary features define the distinct developmental period between
adolescence and young adulthood known as emerging adulthood. The five domains
(Arnett, 2004b) are: (1) age of identity explorations (primarily in relationships and work),
(2) age of instability, (3) utmost self-focused time of life, (4) age of feeling in-between
(no longer an adolescent, but not yet an adult), and (5) the age of possibilities (unmatched
optimism and number of opportunities available).
Age of identity explorations. Previous developmental psychologists have largely
considered identity exploration as a major factor during adolescence (Erikson, 1950).
Identity exploration is still an important step before reaching adulthood, but it is now
taking place during emerging adulthood in addition to adolescence. During emerging
adulthood, unlike any other period in life, young people are able to explore various
possibilities for their lives in domains such as love and work before they are forced to
make more permanent choices in their adult lives. Emerging adults typically do not live
at home anymore, but have not yet began long-term commitments such as job, marriage,
and parenthood. Because these individuals are not bound to their parents, but are not yet
entrenched in adult roles, they have the extraordinary opportunity to explore different
ways of living through friendships, romantic relationships, and work. While identity
formation has typically been considered to mostly occur during adolescence, recent
research confirms that it is actually occurring later than it did previously (Arnett, 2004b).
It makes sense that during the time prior to taking on adult roles, emerging adults are
forming a more solid identity that will help empower them to choose their long-term
13
romantic partner, decide what type of education to pursue, and make difficult decisions
on career paths. During this period emerging adults are constructing their personal
worldview, beliefs, and values separate from that of their parents or guardians.
Age of instability. Emerging adulthood is considered a time of instability because
of the multitude of changes that take place during this time with respect to work, love,
and education. For example, the rate of moving during emerging adulthood is reflective
of the intense changes occurring during this period (Arnett & Tanner, 2006). Most
individuals move out of the parents’ home around 18 or 19 years of age, either to move
out into a college dormitory, or to achieve independence (Goldscheider & Goldscheider,
1999). After living in a dorm, many students move to apartments with friends, then often
cohabitate at some point during their 20s, and increasingly move back to their parents’
home when relationships dissolve or jobs are difficult to find (Bumpass & Lu, 2000).
Residential change is used here to illustrate the extent of the instability that characterizes
the period between the late teens and mid-twenties. However, instability can be found in
a wide variety of domains during this period (e.g. work, romantic relationships).
Self-focused age. Emerging adults do not yet have adult obligations, and thus
experience significant autonomy. It is important to differentiate being self-focused from
being self-centered. Emerging adults are not self-centered; in fact they typically have
high quality relationships with their parents and are very empathetic (Arnett, 2004b).
Children and adolescents typically live with their parents and are obligated to follow their
rules and routines; young people must also follow rules imposed by institutions such as
schools. Having resolved conflicts typical of adolescence, after age 30 many individuals
14
are married, have children, and careers that do not allow them the same kind of freedom
from obligations they had during emerging adulthood. It is for these reasons that
emerging adulthood allows individuals to focus on the self and develop a sense of who
they are. While emerging adulthood is debatably the freest time of life, it is also a time
where individuals spend the most time alone (with the exception of the elderly) (Larson,
1990). Consequently it is viewed more as a transitional time on the path to adulthood.
Age of feeling in-between. Emerging adulthood is considered a time of feeling in-
between; it fits the concept of not yet being a young adult, yet having moved past
adolescence. Emerging adults consistently state the following as the main criteria for
having reached adulthood (Arnett, 1998, 2004b; Facio, Resett, Micocci, & Mistrorigo,
2007):
• Accepting responsibility for the self,
• Making independent decisions,
• Becoming financially self-sufficient.
While emerging adults are beginning to meet these criteria, they do not yet fully
possess these qualities. It is important to note that the main criteria stated for having
reached adulthood are not milestones that take place during a specified time like getting
married or having a child, but instead are gradually reached indicators.
Age of possibilities. Emerging adulthood can also be considered a time of
possibilities. It is a time a possibilities for two distinct reasons. First, the period from the
late teens to early 20s is characterized by great optimism (Hornblower, 1997) and hopes
for the self. Individuals have not yet experienced failures so it is easier to imagine a
15
future where they are happy and fulfilled than it might be for older adults. Even with the
knowledge that 50 percent of marriages end in divorce, and knowing that many people
are not happy with their current job situations, emerging adults hold strong to the belief
that their future will be brighter (Arnett & Tanner, 2006). The second reason is that this
period might be a way for emerging adults to escape troubled or difficult home-life
experiences, and move on to a more independent life in which they can essentially “start
over.” This is the first period in the emerging adult’s life that they are free from their
parents. If the parents fought frequently, had mental illness, or any number of other
problems, emerging adulthood provides an opportunity for this individual to explore
alternative possibilities and liberate them from a dysfunctional environment. For those
who come from relatively happy families, emerging adulthood is still a time to live
independently and forge their identity under new circumstances.
The heterogeneity of emerging adulthood
While emerging adulthood is a period characterized by exploration, instability,
being self-focused, feeling in-between, and having many possibilities, it is also the period
where variance in lifestyle is the greatest (Arnett, 2000a). Every population or subgroup,
and even every individual, experiences this period differently. Not all emerging adults
are at the same place in life; some are in school, some have jobs, some are in school and
have a job, some have children, some are single, some are married, some live alone, some
live at home, some live with roommates. The reasons for this diversity are complex;
these differences reflect the lack of societal norms during this specific period and the
16
newfound freedom these individuals are experiencing. Society is comprised of social
norms that regulate the behaviors of individuals. What makes emerging adulthood so
different from other periods of life is the diversity that exists during this time. For
example, adolescents are expected to not be engaged in serious romantic relationships,
and those over 30 are expected to be long-term romantic relationships. Yet, emerging
adults are between boundaries; being in a long-term relationship is acceptable, while not
being in one is just as accepted. This freedom from strict norms allows emerging
adulthood to be a particularly exciting yet volatile period of life where individuals are
free to explore and discover more about which path they will choose in life.
For decades it was believed that adolescents transitioned directly into young
adulthood. Only in the last 20 years has the period roughly from 18-25 been recognized
as a distinct developmental period. Some might question the validity of calling emerging
adulthood a discrete period with certain shared features (Arnett & Tanner, 2006).
However, the heterogeneity of this period should not deter researchers from accepting the
distinctiveness of this time of life. It is partially because of this great diversity that
emerging adulthood is a distinct period. It is indisputable that the median age of marriage
has risen, more individuals than ever are pursuing higher education, and this period
represents the most changes in residence than any other period in life. Thus emerging
adulthood merits recognition as a distinct period and warrants deeper exploration.
17
Measuring emerging adulthood
Emerging adulthood has been measured in various ways since the term was first
coined (Arnett, 2000b, 2004a, 2004b). Some researchers have looked simply at age
range (Chassin et al., 2002; Luyckx, Goossens, & Soenens, 2006; Shulman & Ben-Artzi,
2003); however, this approach might be limited due to variation in maturity, schooling
and other factors. This way of thinking looks at a specific period of life, but does not
necessarily account for the variation among individuals during this time or the fact that
the time period of emerging adulthood might not be the same for every individual. By
studying emerging adulthood in such a way, it is not possible to determine how deeply
the individual is entrenched in this period; importantly, an age-focused approach might
miss key developmental differences between individuals.
Others have looked at life developmental milestones to determine whether an
individual has reached adulthood (Griffin, 1993; McRobbie, 1991; Stam, Hartman,
Deurloo, Groothoff, & Grootenhuis, 2006). Typical markers of entrance into adult status
include such events as establishing a stable residence, completing school, selecting a
permanent career, and marriage or a long-term stable commitment to a significant other.
It is often assumed that the transition from adolescence to adulthood is linear, but instead
of conceptualizing adolescence as a static classification that individuals outgrow, it is
possible to instead think of it as a growth process that shapes individuals during their
lives; perhaps involving several iterations including gradually accepting responsibility for
the self, making independent decisions, and becoming financially independent (Arnett,
1998, 2004a, 2004b; Facio et al., 2007).
18
Some have used or adapted measures such as the Erikson psychosocial inventory
scale (Rosenthal, Gurney, & Moore, 1981) as a global multidimensional construct
composed of indicators of autonomy, identity, industry, initiative, and intimacy; all of
which have been regarded as requirements for a successful transition to adulthood.
However this scale was built using Erikson’s developmental stages (Erikson, 1950,
1959), where a clear resolution of conflicts is necessary to transition into the subsequent
stage of development. Erikson postulated that there are eight stages through which a
healthy individual passes through between infancy and late adulthood. Each stage is
characterized by a specific conflict that needs to be confronted and resolved. For
example, in the first two years, the infant is confronted with the conflict of trust vs.
mistrust. Specifically the scale itself was designed to determine whether respondents
have resolved conflicts associated with the first six psychological stages developed by
Erikson, and thus does not directly capture emerging adulthood as opposed to
adolescence (Burns & Dunlop, 1998).
Others have utilized self-reports of whether or not persons consider themselves to
be an adult (Mayseless & Scharf, 2003). Such self-report data have been useful in better
defining what adulthood means subjectively to individuals. For example, one study
(Arnett, 1997) had youth to respond to a series of 40 items and indicate whether these
items were “necessary for adulthood”. One Israeli study (Mayseless & Scharf, 2003)
used a similar procedure where participants indicated the extent to which each of forty-
seven items was considered necessary for the individual to be considered an adult. This
study further demonstrated that respondents regarded psychological attributes (i.e.
19
accepting responsibility, deciding on personal beliefs by oneself, and forming equal-adult
relationships with parents) as being the most important markers of adulthood. While
these studies are valuable foundational investigations to determine the subjective factors
that individuals find important for the transition to adulthood, they do not describe the
psychosocial issues or processes that are involved as youth transition through emerging
adulthood (Arnett, 1997). Only recently have people looked at a measure developed to
directly tap the psychosocial issues associated with emerging adulthood, the Inventory of
the Dimensions of Emerging Adulthood (Reifman, Arnett, & Colwell, 2007).
Measuring emerging adulthood: The Inventory of Dimensions of Emerging
Adulthood (IDEA). Currently there is only one measure of emerging adulthood, the
Inventory of Dimensions of Emerging Adulthood, the IDEA (Reifman et al., 2007). The
IDEA is a 31-item Likert-type scale. Respondents are instructed to: “Please think about
this time in your life. When we say ‘this time,’ we mean what is going on now, plus what
has gone on in the last few years, plus what you think your life will be like in the next
few years. Think about a 5-year period of time, with right now in the middle. For each
question below, mark the box that best describes this time in your life. Be sure to put only
one check mark per line.” Items are measured on the following 4-point Likert type scale:
1 (definitely not), 2 (probably not), 3 (maybe), and 4 (definitely yes). The scale is
composed of six subscales: identity exploration (7 items), experimentation/possibilities (5
items), negativity/instability (7 items), other-focused (3 items), self-focused (6 items), and
felling “in-between” (3 items). See Table 1 below. Reifman, Arnett, and Colwell (2007)
investigated the psychometric properties of the 31-item IDEA. Samples were based out
20
of a large university in Texas. Participants were from an undergraduate research methods
course, and friends, family and acquaintances of these students. The authors found that
using both exploratory and confirmatory analyses, the five-subscale structure was largely
supported. Results indicated an additional subscale, other-focused, as a reciprocal to the
self-focused scale. Internal consistency (alpha) reliability coefficients were between .70
and .85, and test-retest reliability correlations (over a one month interval) were also high
(.64 to .76) except for one scale, feeling “in-between”, which had a score of .37.
21
Table 1
IDEA items and dimensions
Dimension/IDEA item
Experimentation/Possibilities
Time of many possibilities?
Time of exploration?
Time of experimentation?
Time of open choices?
Time of trying out new things?
Identity Exploration
Time of finding out who you are?
Time of seperating from parents?
Time of defining yourslef?
Time of planning for the future?
Time of seeking a sense of meaning?
Time of deciding on your own beliefs and values?
Time of learning to think for yourself?
Negativity/Instability
Time of confusion?
Time of feeling restricted?
Time of feeling stressed out?
Time of instability?
Time of unpredictability?
Time of many worries?
Time of high pressure?
Other-focused
Time of settling down?
Time of responsibility for others?
Time of commitments to others?
Self-focused
Time of personal freedom?
Time of responsibility for yourself?
Time of optimism?
Time of independence?
Time of focusing on yourself?
Time of self-sufficiency?
Feeling "In-Between"
Time of feeling adult in some ways but not in others?
Time of gradually becoming an adult?
Time of being not sure whether you have reached full adulthood?
22
Previous Research on the IDEA. Electronic databases (Google Scholar, PubMed,
and Ovid) were searched to identify articles that used the IDEA (accessed February 1,
2011 to March 5, 2011). The databases were searched by using the terms “The Inventory
of Dimensions of Emerging Adulthood,” “Inventory of Dimensions of Emerging
Adulthood,” and “Emerging Adulthood IDEA.” Searches were refined to include articles
published between January 2007 and February 2011. The year 2007 was selected as a
starting point since it was the year the seminal article about the IDEA was published
(Reifman et al., 2007). In addition to computer-assisted searches, bibliographies of
acquired articles were scanned to obtain additional studies. Using these inclusion and
exclusion criteria, the number of empirical paper web pages to examine was 21 (Google
Scholar) and 20 (across PubMed and Ovid) and the total number of non-overlapping
results was 22.
A total of 13 of these web pages were from student theses or dissertations (Duris,
2009; Gollubits, 2010; Gorski, 2009; Hager, 2009; Herrera, 2009; Kreyszig, 2007;
Macek, Bej ek, & Vaní ková, 2007; Musante, 2010; Schönhart, 2009; Sirsch, Dreher,
Mayr, & Willinger, 2009; Supper, 2009; Uranitsch, 2008; Williams, 2007), one was from
a compilation of communication policy and research studies from a conference in
Australia (Williamson, 2007); and eight were peer reviewed articles (Arias & Hernandez,
2007; Arnett, 2006, 2007a; Atak & Çok, 2008; Buhl & Lanz, 2007; Facio et al., 2007;
Horowitz & Bromnick, 2007; Marzana, Pérez-Acosta, Marta, & González, 2010). In
addition, of the 22 results, seven were dissertations in German language, and one was in a
Spanish language peer reviewed journal article. Three of the eight peer reviewed articles
23
were from a special issue (Arnett, 2007b) on emerging adulthood in the Journal of
Adolescent Research. The results included research from Spain, Mexico, Turkey,
Australia, Germany, Canada, and the Czech Republic. None of the 22 studies used the
full IDEA in English. However, one article used a translated IDEA (Turkish; (Atak &
Çok, 2008), and others based their items on the original IDEA and then translated the
new scale into a different language (Arias & Hernandez, 2007; Macek et al., 2007; Sirsch
et al., 2009). Some studies used only one subscale, or a 5-item reduced scale (Herrera,
2009; Holahan, Pasch, & Steinhardt, 2009; Musante, 2010). Herrera (2009) used five
items from three subscales (feeling in between, experimentation, and identity exploration)
as demographic items to demonstrate that the sample was indeed experiencing emerging
adulthood. Musante (2010) used only the Negativity/Instability subscale as she was
interested in studying family predictors of negative instability in adopted emerging
adults. One study (Schwartz, Côté, & Arnett, 2005) examined agency traits, one’s sense
of controlling their own actions, during emerging adulthood in different ethnicities and
found few difference across groups. The study revealed that emerging adults employ
agentic abilities to variable degrees, and that higher levels of agency appear to be related
to exploration and flexible commitment, but was negatively related to avoidance.
The psychometric characteristics of the IDEA were evaluated in a Turkish
sample, adapting the measure to the Turkish language and culture (Atak & Çok, 2008).
The full 31-item scale was utilized. The authors investigated the IDEA’s test-retest
(three-week inteval) and split-half reliability in a sample of university students in Turkey.
A factor analysis was conducted in a sample of high school students, college students,
24
graduate students, and research assistants. This factor analysis (principal components
analysis followed by varimax rotation) revealed a three-factor structure rather than the
six-factor structure found in the original evaluation of the measure (Reifman et al., 2007).
Eleven items were dropped from the original 31. The authors named the three factors
“Negativity/Instability,” “Exploration/Feeling in between,” and “”Experimentation/Self
focused.” Test-retest reliability coefficients were 0.78 for the first factor, 0.76 for the
second, 0.82 for the third, and 0.81 for the total of the scale. The split-half reliability
coefficients were also acceptable, as the first half Spearman-Brown reliability coefficient
was 0.58, the second half’s was 0.65, and the full scale’s, 0.69. Until now, the IDEA has
never been examined in an at-risk population, continuation high school students.
Health Risks of Drug Use Emphasized in Continuation High School Youth
Alternative high schools in California (i.e. continuation high schools) were
created in 1919 to fulfill a mandate that all youth receive part-time education until they
reach adulthood (California Educational Code Section 48400; established in 1919).
Students who are unable to remain in regular high schools for functional reasons such as
credit deficiencies, poor grades, substance misuse, or delinquency, are transferred to the
continuation high school system (Grunbaum et al., 2000).
Continuation high school students are more likely than regular high school
students to have smoked cigarettes, drank alcohol, and used marijuana or other drugs in
the last 30 days (Grunbaum et al., 2000; Sussman, Dent, & Leu, 2000; Sussman, Dent, &
Stacy, 2002; Sussman et al., 1995; Weller et al., 1999). In general, teens who exhibit
25
precocious development (Krohn, Lizotte, & Perez, 1997) or use drugs during adolescence
are at a particularly high risk for alcohol and drug use during the emerging adulthood
period (Brook, Adams, Balka, & Johnson, 2002; Newcomb & Bentler, 1988). One study
found that the prevalence of daily use of cigarettes, alcohol, marijuana and other illicit
substances was 3.9, 2.7, 3.7, and 2.2 times higher, respectively, in one Southern
California continuation high school compared to a population of students attending
regular public high schools in the same region (de Moor et al., 1994). Striking difference
in smoking rates exists between college-bound and non-college-bound high school
seniors (Johnston et al., 2009). In 2009, smoking half a pack or more of cigarettes per
day was found to be three times higher in non-college-bound 12
th
graders compared to
their college-bound counterparts (11.2% vs. 3.6%). Continuation high school youth are
less likely to attend college. One study showed that nearly a quarter of continuation high
school youth fall into the clinical range for alcohol use, and that the majority of these
youth and their families had been to counseling or have suffered from psychiatric
hospitalizations for drug abuse and psychological problems (Franklin & Streeter, 1995).
Emerging Adulthood and Drug Use in Continuation High School Students
Perhaps due to specific characteristics such as feeing in-between, instability,
exploring the identity, and the lack of prescribed behavioral norms (Arnett, 2005)
emerging adults exhibit the highest prevalence of drug use in comparison to individuals
in other periods of life (YRBSS, 2009).
26
Although, high school students are typically considered adolescents, some
evidence demonstrates that continuation high school students reach emerging adulthood
at a younger age (Sussman, 2010). Premature or precocious transitions into adulthood
can have negative implications for drug use (Krohn et al., 1997). The timing, order, and
success of transitions into adulthood may affect the likelihood of beginning, continuing,
or escalating deviant behaviors (Newcomb & Bentler, 1988; Sampson & Laub, 1995).
Some have found that only about 40 to 50 percent of the population follow a “normative
sequence of transitions” into adulthood (Rindfuss, Swicegood, & Rosenfeld, 1987).
What happens to the rest? Specifically, at-risk youth exhibit an increased probability of
encountering disorderly transitions into adulthood (Krohn et al., 1997).
Continuation high school students are more likely to experience the demands of
adult life sooner than youth in a more normative trajectory (Sussman, 2010). These
youth have an augmented likelihood of having children at an earlier age, dropping out of
school, getting married, and working at low-paying jobs. These same individuals are also
more likely to exhibit unconventional attitudes, exhibit a lack of emotional self-control,
experience interpersonal difficulties, and have unplanned pregnancies (Brook, Whiteman,
Cohen, Shapiro, & Balka, 1995).
It is possible that instead of passing though emerging adulthood, continuation
high school students move directly from adolescence to adulthood. While this is
conceivable theoretically, the evidence does not point in this direction. If continuation
high school students passed straight into adult status, certain characteristics should be
present in this demographic. For example, continuation high school students should have
27
found a stable job course, they should focus on others and the self, they should most
likely be married, they should make independent decisions, and they should feel stability,
among other characteristics (Arnett, 2004b). Instead, it is conceivable that continuation
high school students, and other non-normative youth, experience a difficult transition into
emerging adulthood compared to normative-trajectory peers (Sussman, 2010). Sussman
found that in a sample of continuation high school students a number of characteristics
supported the proposition that these youth reach emerging adulthood at an earlier age.
The frequency of alcohol and drug use was high in this sample, educational status was
indicative of a population still in evolution, employment data reflected wanting to take on
different jobs later; all indicative of a sample in transition.
Evidence indicates that difficulties in transition to adult roles is associated with
drug use at an early age (Newcomb & Bentler, 1988; Rohrbach, Sussman, Dent, & Sun,
2005). Disrupted family formation and career development is positively associated with
higher crime and drug use (Thornberry, 1987). Precocious transitions into adult status,
such as teenage pregnancies, increase the chances of marital instability and thus increase
the probability of continued drug use (Furstenberg, Brooks-Gunn, & Morgan, 1989;
Hayes, 1987). As precocious transitions into adult roles might be detrimental to normal
functioning in society, it might be important for drug prevention programs to focus on
youth who are the highest risk for drug abuse.
28
Introduction to the Dissertation Studies
The introduction of this dissertation proposal identified adolescent drug and
alcohol use as a significant health problem and emphasized cause for concern in the
current emerging adulthood drug and alcohol use trends in continuation high school
students in the United States. The relatively new and burgeoning construct of emerging
adulthood was offered as an innovative avenue to investigate additional variance in
youth’s drug and alcohol use. Initial theory and empirical research on emerging
adulthood and substance use was presented.
It should be recognized that very few studies, with some noteworthy exceptions
have effectively studied drug use etiologies longitudinally (D. B. Kandel & Chen, 2000;
Newcomb & Bentler, 1988). The majority of alcohol and drug use studies have been
relegated to using either cross-sectional or two-wave research designs (von Sydow, Lieb,
Pfister, Höfler, & Wittchen, 2002). Longitudinal research designs are important to be
able to identify time-course issues in alcohol and drug use. This type of research design
allows the clarification of issues such as normative vs. non-normative developmental
trajectories, and identification of the most common intra-individual trajectories.
Longitudinal research designs allow for more precise discernment of the risk and
protective factors that are operating in the initiation and continuation of drug and alcohol
use.
This dissertation proposal includes two independent studies that aim to elucidate
the dimensions of emerging adulthood and how they might operate in continuation high
school students in regards to drug and alcohol use behaviors. Study I tests the
29
psychometric properties of the Inventory of Dimensions of Emerging Adulthood
(Reifman et al., 2007) in a sample of continuation high school students and is used to
inform the use of this measure in Study II. Study II examines the predictive influence of
emerging adulthood trajectory on future demographic, and risk behaviors in continuation
high school students. These studies will be the first to examine the utility of the
emerging adulthood construct in continuation high school drug and alcohol use etiology,
and will illuminate the underlying mechanisms that may further inform continuation high
school drug prevention programming in the United States.
Project Toward No Drug Abuse (TND) proposed the IDEA to explore more
directly the dimensions of emerging adulthood among continuation high school youth.
This dissertation will attempt to explore emerging adulthood in this population is
different than the population the measure was originally conceived for, as well as how
successful transitions through emerging adulthood might affect subsequent drug and
alcohol use in continuation high school students. Hence, it is worthwhile to explore to
the relationship emerging adulthood and drug and alcohol use in continuation high school
students.
30
CHAPTER 2: STUDY 1: EVALUATION OF THE PSYCHOMETRIC
PROPERTIES OF THE INVENTORY OF DIMENSIONS OF EMERGING
ADULTHOOD IN A SAMPLE OF CONTINUATION HIGH SCHOOL
STUDENTS
Chapter 2 Abstract
It is now presumed that youth do not move directly from adolescence to
adulthood, but rather pass through a transitional period, “emerging adulthood.” The
Inventory of the Dimensions of Emerging Adulthood (IDEA) is a self-report instrument
developed to examine the attributes of this period. In the present study, a 21-item version
of the IDEA was administered to a sample of 1676 “at risk” continuation (alternative)
high school students in Southern California. Principal component factor analysis with
orthogonal rotation revealed three factors. Overall, the measure demonstrated high
internal consistency. Construct validity analyses indicated that the measure was
correlated with demographics, risk behaviors, and psychological measures. We conclude
that the IDEA is a useful instrument for measuring emerging adulthood in at-risk
populations.
Introduction
Emerging adulthood, the developmental period between adolescence and young
adulthood, is a distinct period demographically, subjectively, and in terms of identity
exploration (Arnett, 2000b). In the course of this life transition, roughly between 18 to
25 years of age, individuals achieve relative autonomy from guardians, such as parents,
and experience shifts in social roles and normative expectations for their behavior.
Emerging adults are typically free from the dependency that characterized childhood (e.g.
31
parents and teachers), yet are not burdened with the responsibilities of adulthood. This
freedom allows emerging adults the opportunity to explore diverse potential life
directions. During this period, more than any other stage of life, the near-future is
uncertain, and individuals are making a variety of life decisions in terms of their
education, work, romance and worldviews (Arnett & Tanner, 2006).
Five distinct dimensions of emerging adulthood have been proposed (Arnett,
2004a, 2004b): the age of identity explorations, the age of feeling in-between, the age of
possibilities, the self-focused age, and the age of instability.
Present Study
Given the limited generalizability of the few psychometric studies of the IDEA, it
is important to examine the reliability and validity of the scale in other population
subgroups. The current study provides a psychometric evaluation of the IDEA in an
emerging adult and predominantly Latino sample of “at risk” older adolescents. The
aims of the study were to:
1. Investigate differences in how the emerging adulthood period is experienced by a non-
white population.
2. Explore the reliability of the emerging adulthood construct in a non-white population.
3. Examine the factor structure of the IDEA using:
a) Confirmatory factor analysis (CFA)
b) Exploratory factor analysis (EFA)
32
4. Explore the predictive validity of the IDEA by examining relations of each factor with
variables that might indicate emerging adulthood transitions, such as demographic
characteristics, risk behaviors, and indicators of psychological maturation.
Method
Participants and procedures
Students were recruited from 24 continuation high schools in Southern California
as part of a larger drug abuse prevention trial (Sussman, Sun, Rohrbach, & Spruijt-Metz,
under review). Schools were recruited as a convenience sample based on student
ethnicity, number of students, and proximity to project headquarters. At least two
classrooms in each school were selected to participate. Students from the selected
classrooms were first asked to complete student information sheet (SIS) that included
their parent’s names, phone number, and student signature. Students over 18 years of age
filled out the forms on their own and were consented at that time. Students under the age
of 18 were asked to bring parent consent forms home. If no parent consent form was
returned, the information on the student assent forms was used to call homes and get
verbal parent consent. If verbal consent was obtained, students were asked to fill out the
SIS at the pretest. For any others that did not have SIS sheets filled out, we attempted to
get the information at pretest from the student.
Of the 2397 students enrolled in the selected classes, 1694 (70.7%) were
consented to participate in the study. Of these, 1676 students completed the pretest
survey. A close-ended, self-report questionnaire was administered to students at pretest.
33
If a student was absent during the data collection day, an absentee packet was left with
instructions. The university’s Institutional Review Board approved all study procedures.
The present sample consisted of youth aged an average of 16.8 years (SD = .90)
who were majority male (57.8%). The sample was 64.3% Latino, 13.3% mixed ethnicity,
11.0% White, 6.3% other ethnicity (not specified), and 5.1% African American. Half
lived with both parents, only 17.5% mostly spoke a language other than English at home,
and approximately 49.3% and 55.9% of youths’ fathers and mothers, respectively, had at
least completed high school.
Survey Measures
Emerging Adulthood. The original IDEA (Reifman et al., 2007) consists of 31
items; however, the reduced measure used in this study consisted of 21 items chosen to
reflect four dimensions of primary interest (Reifman, personal communication, May 27,
2006): a time of identity exploration, experimentation/possibilities, self-focus, and
feeling in-between. Not included were the other-focus and negativity/instability scales.
Other-focus is not one of the five focal dimensions of emerging adulthood; rather it is an
“extra subscale” that enables investigators to see if low self-focus is correlated with high
other-focus (and vice-versa). Negativity/Instability is one of the focal dimensions, but
we eliminated it to avoid content overlap with the stress and anxiety measures included in
the survey.
Using the exact prompt developed by Reifman for the IDEA, respondents were
instructed to: “Please think about this time in your life. When we say ‘this time,’ we
34
mean what is going on now, plus what has gone on in the last few years, plus what you
think your life will be like in the next few years. Think about a 5-year period of time,
with right now in the middle. For each question below, mark the box that best describes
this time in your life. Be sure to put only one check mark per line.” Responses were
measured on the following 4-point Likert type scale: 1 (definitely not), 2 (probably not), 3
(maybe), and 4 (definitely yes). Five items measured experimentation/possibilities, such
as “time of many possibilities?” and “time of exploration?” Six items measured being
self-focused, such as “time of personal freedom?” and “time of independence?” Also,
seven items were used to measure identity exploration, such as “time of separating from
parents?” and “time of defining yourself?” Finally, three items measured feeling in-
between, such as “time of feeling adult in some ways but not others?” and “time of
gradually becoming an adult?” All items were the same as the original measure except for
one from the self-focus scale which was changed from “time of self-sufficiency?” to
“time of providing for yourself without the help of others?” to make the item easier to
understand. See Table 1 for a complete listing of items and proposed subscales.
Demographics and measures regressed on the IDEA and its subscales. Various
categories of correlates were selected to look at their different relations with the IDEA
factors. Certain demographic variables (i.e. job status, marital status, parenthood) were
selected as they might provide concrete markers of the transition to young adulthood.
Control variables included age (in years), gender, and ethnicity (indicator coded as
Latino/Hispanic, African American/ Black, White/Caucasian, mixed, or other).
35
Participants were also asked their job status (0 = no job, 1 = have a job), whether they
were married (0 = no, 1 = yes), and whether they were a parent (0 = no, 1 = yes).
Risk Behaviors. Risk behaviors may indicate difficulties with resolving the
emerging adulthood process. Past thirty day drug use was assessed for cigarettes,
alcohol, marijuana and hard drugs (cocaine, inhalants, ecstasy, stimulants or other hard
drugs) using four items, such as “In the past 30 days, how many days did you smoke
cigarettes?” (Alpha = 0.70).
Engagement coping, stress, and decision-making confidence/avoidance. Other
such variables as decision making, stress and coping were introduced as possible
correlates of the IDEA processes as indicators of a certain type of psychological
maturation. Engagement coping was measured using a 6-item scale, with the question
stem being “The following are things that people may do when they have a problem at
school or at home. Put an ‘X’ in the box that shows how much you do each thing when
you have a problem.” Examples of engagement coping included “I talk to my
mother/father” and “I think about the choices before I do anything” (1 = never to 5 =
always; alpha = 0.85). Stress was measured using a 4-item variation of the perceived
stress scale (Cohen, 1988; Cohen, Kamarck, & Mermelstein, 1983); one sample item
was, “During the last month have you ever been upset because of something that
happened unexpectedly”(alpha = 0.87). Decision making was measured with the
decision-making confidence and decision-making avoidance scale (Commendador, 2007;
Tuinstra, Van Sonderen, Groothoff, Van den Heuvel, & Post, 2000). The 3-item
decision-making avoidance scale included 4-point Likert response categories from never
36
to always. A sample item includes, “When faced with a decision, I go along with what
others suggest” (alpha = 0.73). Decision-making confidence was measured with a 3-item
scale with items such as “I like to make decisions myself” (alpha = 0.73).
Statistical analyses
First, confirmatory factor analysis (CFA) with maximum likelihood was
employed to assess how well the data fit the four-factor model using Mplus (L. Muthén &
Muthén, 2007). Each subscale represented a latent variable and each item was an
indicator variable. As the CFA indicated poor model fit, as is described below, and
because exploratory factor analysis (EFA) is often used in scale development (Hurley et
al., 1997), the internal structure of the IDEA was subsequently examined using two
complimentary techniques: EFA and an analysis of internal consistency. An EFA
analysis was carried out by means of principal component analysis (PCA). Using the
PCA, only factors with eigenvalues > 1.0 were retained. Next, the scree plot was visually
inspected. This procedure was paired with a varimax (orthogonal) rotation of the factor
structure to achieve a simpler structure with greater interpretability. Exploratory factor
analysis methods use certain “rules of thumb,” with factor loading cutoff criteria ranging
from .30 to .55, to establish a solid factor loading coefficient (Swisher, Beckstead, &
Bebeau, 2004); we used a cutoff value of .40 in this study. Internal consistency was
examined by computing Cronbach’s alpha for the entire measure and each subscale.
Regressions (mixed models and logistic) were used to examine associations
between subscales while controlling for age, gender, and ethnicity. School (which was
37
the unit of randomization) was considered a random effect. This allows for both the
statistical accounting of intraclass correlation within clustered units (school) on computed
significance levels, and for generalization of the findings to a larger sample. Items such
as demographic, behavioral, and psychosocial variables were considered to explore the
construct validity of the IDEA in this sample. All analyses other than the CFA were
performed using SAS version 9.2 (SAS, 2008).
Because factor structures often differ by gender, a test for factorial invariance
across males and females using the factor structure determined by the EFA was
conducted. Factor loadings, variances, and covariances were constrained to be equal
across groups and then model fit was evaluated.
Results
Confirmatory Factor Analysis (CFA)
The CFA model contained the following specification: (1) four dimensions
corresponding to the theoretical model of emerging adulthood (the four factors from
Reifman et al., 2007); and (2) 21 observed variables. The model was identified according
to appropriate rules for CFA (Bollen, 1989). The chi-square statistic for the model was
1861 with 183 degrees of freedom (p < .0001), suggesting poor model fit. As further
indication of poor model fit, the comparative fit index (CFI) was below 0.9, and root
mean squared error of approximation (RMSEA) was above 0.05.
38
Figure 4. IDEA factor structure model per Reifman, Arnett & Colwell (2007).
39
Principal Components Analyses (PCA) and Internal Consistency
The PCA indicated three factors. Before orthogonal rotation, Factor 1 accounted
for 41.1% of the variance (eigenvalue = 8.64), Factor 2 accounted for 7.1% of the
variance (eigenvalue = 1.49), and Factor 3 accounted for 5.4% (eigenvalue = 1.12). After
an orthogonal rotation, the three factors showed a less striking difference in eigenvalues
(5.10, 3.78, and 2.37). As shown in table 2, every item loaded above .43 on at least one
factor. Only two items loaded at .40 or greater on more than one factor: item 9, “Time of
open choices?” on factors 1 and 2, and item 12 “Time of focusing on yourself?” on
factors 1 and 3. Factor 1, comprised of 11 items, appears to represent Identity
Exploration. Factor 2, containing 7 items, could be described as
Experimentation/Possibilities. Factor 3, composed of 3 items, could be described as
Independence or Self-Focus. Similarly, visual inspection of the scree test indicated that
all three factors should be retained.
40
Table 2
IDEA
Subscale
IDEA item (Domain) Factor 1:
Identity
Exploration
Factor 2:
Experimentati
on/Possibilitie
s
Factor 3:
Indepedence
E/P 1. Time of many possibilities? 0.37 0.67 -0.05
E/P 2. Time of exploration? 0.21 0.79 0.06
E/P 3. Time of experimentation? 0.09 0.79 0.09
E/P 9. Time of open choices? 0.44 0.49 0.34
E/P 11. Time of trying new things? 0.24 0.52 0.39
SF 4. Time of personal freedom? 0.16 0.58 0.40
SF 5. Time of responsibility for yourself? 0.52 0.36 0.21
SF 6. Time of optimism? 0.31 0.60 0.13
SF 8. Time of independence? 0.38 0.36 0.53
SF 12. Time of focusing on yourself? 0.44 0.25 0.41
SF
10. Time of providing for yourself without
the help of others? 0.29 0.10 0.67
IE 7. Time of finding out who you are? 0.54 0.37 0.16
IE 13. Time of separating from parents? 0.12 0.04 0.76
IE 14. Time of defining yourself? 0.59 0.30 0.37
IE 15. Time of planning for the future? 0.71 0.23 0.14
IE 16. Time of seeking a sense of meaning? 0.69 0.24 0.19
IE
17. Time of deciding on your own beliefs
and values? 0.71 0.25 0.20
IE 18. Time of learning to think for yourself? 0.73 0.28 0.12
F-IB
19. Time of feeling adult in some ways but
not in others? 0.68 0.19 0.14
F-IB 20. Time of gradually becoming an adult? 0.72 0.14 0.21
F-IB
21. Time of being not sure whether you
have reached full adulthood? 0.43 0.03 0.24
Derived factors
Note : Item factor loadings > .40 are boldfaced. If item loaded >.40 on more than one factor, higher
loading was boldfaced.
Four subscales were assessed from the original IDEA, abbreviations are as follows: E/P =
Experimentation/Possibilities, SF = Self Focused, IE = Identity Exploration, F-IB = Feeing "In-
Between"
PCA factor loadings after orthogonal rotation
41
The Cronbach’s alpha for the entire IDEA was .93 after standardization.
Cronbach’s alphas were computed for each subscale and found to be as follows: Factor 1,
.89; Factor 2, .85; and Factor 3, .64.
Inter-factor Correlations
As expected, the three factors were highly inter-related, with correlations ranging
from .51 to .69 (p’s < .0001; see Table 3).
Exploration of Construct Validity of the IDEA
In ordinary and logistic regression analyses that controlled for age, gender, and
ethnicity, a number of relationships between demographic, health/risk, and psychological
variables, on the one hand, and the IDEA or its subscales, on the other hand, were
significant (see Table 4). In the demographic realm, higher scores on the full IDEA
scale, identity subscale, and experimentation/possibility (E/P) subscale were associated
with lower odds of having a job, whereas higher scores on the independence subscale
were linked to lower odds of being married or cohabiting. Higher full IDEA and E/P
Table 3
Correlations between factors
Subscale Number of items Factor 1 Factor 2 Factor 3
Factor 1 11 -
Factor 2 7 0.69 -
Factor 3 3 0.61 0.51 -
Note: All correlations were p <.0001. Factor 1 = Identity
Exploration, Factor 2 = Experimentation/Possibilities, Factor 3 =
Independence.
42
scores were associated with greater odds of parenthood. The E/P factor (and to a lesser
extent Independence) were related to a variety of health/risk behavior variables. Feeling
that one was experiencing a time of experimentation and possibilities was tied to greater
cigarette smoking and alcohol and marijuana use. Finally, regarding psychosocial
correlates, higher scores on the total IDEA and all subscales went along with higher
reported stress, engagement coping, and decision-making avoidance (as well as less
decision confidence).
43
Table 4
Variable IDEA (full scale)
β or OR(95%CI) p β or OR(95%CI) p β or OR(95%CI) p β or OR(95%CI) p
Job (0 = no job, 1 =
job)
OR(95%CI) =
0.72 (0.54-0.96) <.05
OR(95%CI) =
0.74 (0.56-0.91) <.05
OR(95%CI) =
0.79 (0.63-0.99) <.05
OR(95%CI) =
0.87 (0.72-1.10) 0.15
Married (0 = no, 1 =
yes or no but living
with significant other)
OR(95%CI) =
0.84 (0.50-1.40) 0.50
OR(95%CI) =
0.77 (0.46-1.3) 0.33
OR(95%CI) =
1.24 (0.85-1.81) 0.27
OR(95%CI) =
0.61 (0.42-0.88) <.01
Parent (0 = no, 1=
yes)
OR(95%CI) =
1.51(1.02-2.25) <.05
OR(95%CI) =
1.33 (0.90-1.97) 0.16
OR(95%CI) =
1.82 (1.32-2.51) <.001
OR(95%CI) =
0.86 (0.62-1.18) 0.35
30-day cigarette use
(number of days) 0.02 0.27 0.00 0.87 0.05 <.05 0.03 0.17
30-day alcohol use 0.03 0.24 0.01 0.58 0.07 <.01 0.06 <.05
30-day marijuana use 0.06 <.05 0.00 0.98 0.10 <.0001 0.05 <.05
30-day hard drug use 0.00 0.44 0.00 0.88 0.03 0.17 0.04 0.13
Engagement Coping 0.24 <.0001 0.25 <.0001 0.22 <.0001 0.07 <.01
Stress 0.26 <.0001 0.31 <.0001 0.10 <.05 0.13 <.001
Decision Making
Confidence -0.26 <.0001 -0.26 <.0001 -0.19 <.0001 -0.09 <.0001
Decision Making
Avoidence 0.14 <.0001 0.14 <.0001 0.10 <.0001 0.07 <.01
Regressions of the IDEA full scale and subscales with other variables
Factor 1 (Identity) Factor 2
(Experimentation/Possibilities)
Factor 3 (Independence)
Note: All results are controlling for gender, age, and ethnicity * p < .05
44
Exploration of Factorial Invariance of the IDEA across Genders
To test for factorial invariance between males and females factor loadings,
variances, and covariances were constrained to be equal across groups. Model fit
changed very little from the one group model to the two-group fully constrained model;
both models indicated adequate fit. Model fit indices using the EFA results in a CFA
model for one group were the following: χ
2
= 1825.4, CFI = 0.88, TLI = 0.87, RMSEA =
0.074 [0.071 – 0.077]. Results for the multi-group model were the following: χ
2
male =
1043.8, χ
2
female = 1243.7, CFI = 0.87, TLI = 0.87, RMSEA = 0.075 [0.072 – 0.078].
Thus, it was determined that the factor structure did not vary significantly between males
and females.
Discussion
The aim of the present study was to examine the psychometric properties of the
IDEA (Reifman et al., 2007) when used in a sample of high-risk, older adolescents. An
EFA was conducted that yielded meaningful subscales; three factors accounted for 53.6%
of item variance. Three subscales were identified (identity exploration,
experimentation/possibilities, and independence). It is possible that the differences in
scale structure between our study and that of Reifman and colleagues (2007) can be
explained by the differences in samples used in the two studies. The original scale was
developed for 18-25 year olds and validated in a sample of predominantly White female
college students and non-college, same age, acquaintances (Reifman et al., 2007),
45
whereas the current sample was comprised of continuation high school youth in which
the majority group were 17 year old male Latinos.
There are several reasons why the differences in gender, age, and ethnicity of the
participants in the two studies might be important. As the current sample contained a
higher percentage of males and somewhat younger participants than in Reifman et al.
(2007), the current results could potentially reflect students experiencing emerging
adulthood a bit differently than an older, predominantly female sample would. However
because continuation high school students have had experiences that go along with
emerging/young adulthood (having children, drug use, etc.), it is likely that they would
have entered it earlier. Indeed, while emerging adulthood (Arnett & Tanner, 2006) has
been described as a theory of development from the late teens through the early 20s, the
age guidelines are approximate. Evidence suggests that emerging adulthood is likely
experienced as a transitional period in continuation high school youth (Sussman, 2010).
However differences based on gender, age and ethnicity were not examined specifically,
so these interpretations are purely speculative.
Ethnicity may also explain differences in how emerging adulthood is experienced.
Some minority youth may have a stronger sense of interdependence and responsibility to
their families than non-minority youth (Arnett & Tanner, 2006). Some minority cultures
place more pressure on having their constituents marry and have children at a younger
age (Phinney, 1999). While marriage or having children does not necessarily act as an
indicator of reaching adulthood (Arnett, 1998), it is possible that these individuals will
consider themselves as adults. As such, it is plausible that young people from minority
46
cultures reach typical markers (e.g. marriage, having children) of adulthood earlier and
thus experience a shorter period of emerging adulthood. Also, other data indicate that
while these youth are relatively likely to have children, they also report not being settled
down in terms of education and employment, even at an average of 22 years old
(Sussman, 2010). Thus, at-risk youth from minority cultures may have an extended
period of emerging adulthood that manifests itself differently from other groups. Much
more research is needed to examine this possibility.
Reduced Factor Structure
Scale developers proposed that the shortened IDEA measure was composed of
four dimensions (Reifman et al., 2007); however the factor analysis performed for this
paper suggests that the scale is composed of three subscales (Identity Exploration,
Experimentation/Possibilities, and Independence) when administered to the study
population of late adolescent continuation high school students. Essentially two
subscales (Identity Exploration and Feeling In-Between) were collapsed into one
subscale, newly labeled for the one dimension: Identity Exploration. Only one item was
left out from the original Identity Exploration subscale, and two items were added from
the previously labeled Self Focus scale. The second factor, Experimentation/Possibilities
remains largely unchanged; two items from the Self Focused scale were added to the
initial dimension; thus extending it from five items to seven items. The third factor
represents the main difference between the proposed dimensions (Reifman et al., 2007)
and the current analysis; we labeled this factor Independence. It comprises three items
47
from two distinct proposed dimensions, Self-Focused and Identity Exploration. Changes
from original subscales can be found in Table 2.
Using the new reduced factor structure, “Independence” (time of independence,
providing for oneself, separating from parents) is of particular interest since it differs
most from the original proposed dimensions. It is possible that this factor emerged as a
direct result of the population studied. The items for this new factor were initially from
the original subscales Self Focused and Identity Exploration. Continuation high school
youth might experience emerging adulthood as a time of Independence more so than
mainstream youth. Continuation high school students are in a potentially precarious
situation; they have the option either to get their studies back on track, or drop out of
school and join the workforce. Graduation rates at continuation high school are
significantly lower than rates at regular high school (Sussman, Rohrbach, Skara, & Dent,
2004). It is possible that the “sink or swim” environment of continuation high schools
changes the way these youth experience emerging adulthood. These individuals likely
are in an in-between stage, already experiencing more independence than their regular
school counterparts (Sussman, 2010). Two of the items in the Independence scale
originated from what was originally labeled the Self-Focus scale, indicating that perhaps
continuation high school youth are not mature enough to fully explore and differentiate a
sense of self, but they are already experiencing increased independence and making life-
defining decisions for themselves (previously referred to as “truncated development” by
(Newcomb & Bentler, 1988).
48
The IDEA was found to be internally consistent in the current study, with two
factor-based subscales having Cronbach’s alpha values at or above .85 and the third
subscales, with only three items, having an alpha of .64. The subscales showed
considerable inter-correlations, as was found in the study by Reifman et al. (2007). This
finding is consistent with the proposition that while the subscales measure different
domains, they all reflect the larger conceptualization of emerging adulthood. However, it
is important to use the proposed three-factor structure as it adds considerable information
and differentiation between the various domains that compose emerging adulthood
(specifically between Factor 2 and the others; see Table 4). For example, being married is
associated with a lower level on Factor 1 and Factor 3, but a higher level on Factor 2.
Investigators who wish to utilize the Reifman measure to explore emerging adulthood
concepts among Latino youth may want to attempt a replication of the scale structure
supported by the present analyses.
Although additional work relating the IDEA to other measures is clearly needed,
these preliminary results are promising. Assuming that scoring high on the emerging
adulthood scale indicates that the individual is in the midst of identity
exploration/experimenting, being open to possibilities, and finding gradual independence,
we can make a number of inferences about a population of continuation high school
youth. Scoring low on the IDEA could reflect a number of possibilities. For example, it
could mean that such individuals have “resolved” emerging adulthood and are living
independently (perhaps having reached adulthood sooner than most same-age peers), or it
could mean that they have not yet entered emerging adulthood and are currently living at
49
home with their parents. This scoring is suggestive of a non-linear relation. Of course,
with continuation high school youth, the context is different than with college youth, and
emerging adulthood could be indicative of a rather different developmental context. For
example, getting married among continuation high school youth may indicate
experimenting with life options, as opposed to college-educated youth’s sequence of
marriage being a culminating accomplishment of young-adult development after the
attainment of educational and career goals (Settersten Jr & Ray, 2010).
The regression results provide evidence of construct validity as well as insights
into what continuation high school students might be experiencing. Decision making
avoidance, stress, and engagement coping are all positively related to both the full IDEA
scale and each individual factor. Higher scores on the full IDEA, identity-seeking, and
experimentation/possibilities were associated with not yet having a job. Having a job
might have a different meaning for continuation high school youth than it does for
mainstream high school youth. Perhaps having a job allows these youth to explore more
possibilities through their jobs and to explore the self and what they like. In a sense, the
jobs are distractions for momentary gains, thus avoiding long-term and more meaningful
goals. As has been previously found in another sample (Sussman, 2010), these youth
most likely do not perceive these jobs as careers, and therefore are still situated in
emerging adulthood instead of full-fledged adulthood. Decision making confidence was
negatively related to the full IDEA, and each subscale, while decision making avoidance
and stress were positively related; thus supporting the concept that emerging adulthood is
a period of uncertainty and learning how to develop confidence in how to make decisions
50
that are important for the future. Engagement coping might have been positively related
to emerging adulthood because engagement coping involves being active and holding a
sense of agency which is part of the maturation process (Long, 1998).
As previously discussed, marriage was negatively correlated with the
independence factor, such that being married meant a lower score on independence. This
makes sense as marriage is often associated with a loss of independence since the
individual now must not only act for themselves, but also take into account another
person. The experimentation/possibilities factor was positively related with 30-day
substance use (cigarettes, alcohol, and marijuana). Perhaps youth who score higher on
experimentation are experimenting with various substances among other things.
Limitations
The present study was not without its limitations. First, as with any psychometric
evaluation, replication is necessary. Only cross-sectional data were used, so replication
with longitudinal sampling may be helpful to better understand changes on the IDEA
over time, and the order of precedence between its factors and other variables. Second,
several differences in sample characteristics (age, gender, ethnicity) might explain
differences between the original (Reifman et al., 2007) and current psychometric
evaluations of the IDEA. Third, the present participants were continuation high school
students, so the results of this study cannot be generalized outside of this population.
Fourth, it might be a limitation that the full IDEA scale with all six subscales was not
administered. Lastly, it could be considered a limitation that the subscales are not
51
completely distinct (some cross loadings); however, this also indicates that the subscales
are part of the overarching concept of emerging adulthood.
Conclusion
The IDEA is a useful and practical instrument for measuring emerging adulthood.
The scale possesses acceptable psychometric properties in terms of subscales and internal
consistency. To our knowledge, this is the first evaluation of this scale with a population
of continuation (alternative) high school students. Thus, this instrument can be applied to
research on developmental transitions among at-risk youth. The present psychometric
evaluation of the IDEA in this diverse at-risk, older adolescent population perhaps
suggests a more suitable dimensionality for the measure than originally proposed.
Moreover, additional research should investigate the new dimensionality in diverse
samples of youth who are in the definitive emerging adulthood period (18-25 years).
Finally, future studies should examine test-retest reliability as well as establish the
reliability and validity of this scale for other populations.
52
CHAPTER 3: STUDY 2: THE RELATIONSHIP OF EMERGING ADULTHOOD
TRAJECTORIES TO DRUG USE, AND OTHER BEHAVIORAL OUTCOMES
ACROSS THREE WAVES OF DATA
Chapter 3 Abstract
Introduction: Over the past decade there has been interest in examining the period
between adolescence and young adulthood, emerging adulthood, as a unique
developmental period. This study examined the heterogeneity of the period of emerging
adulthood by identifying trajectories of development across individuals and identifying
how these trajectories differ in terms of future outcomes.
Method: A sample of continuation high school students as part of a larger drug
prevention project were used to identify trajectories of emerging adulthood. First, the 8-
item IDEA measure was tested for factorial invariance across three time points. Factorial
invariance was achieved using a 5-item version. Next, latent growth modeling was used
to identify the overall trajectory of emerging adulthood over time. Next, latent class
growth modeling was used to identify unique developmental trajectories over time from a
larger heterogeneous sample. Lastly, classes were compared in terms of both baseline
and Time 3 correlates.
Results: Three classes were identified. These classes differed in important ways both in
terms of intercept and slope, but also in terms of baseline characteristics. Class 3
appeared to be the most at-risk at Time 1. However, at Time 3, the class appeared to be
very similar in terms of demographic characteristics and risk behaviors.
Discussion: Latent class differs in baseline characteristics but comes together at Time 3.
53
Introduction
Although data on drug use etiologies are high in volume, and longitudinal
analyses of these antecedents are becoming more popular, most studies fail to investigate
how drug and alcohol use change over time and what factors impact these patterns of
change. These studies test only cross-sectional or short-term (two-wave) longitudinal
data (Chassin et al., 2004; Cole & Maxwell, 2003). Non-linear variability, individually-
varying times of observation and auto-regressive (AR) effects (prediction of one time
point based on the previous time point) cannot be sufficiently captured without using
longitudinal methodologies (Cheong, MacKinnon, & Khoo, 2003; Lawrence & Hancock,
1998; Rose, Chassin, Presson, & Sherman, 2000).
Even AR models which are often used to model natural occurrences are limited
by the fact that they assume linear prediction. Typical AR models do not provide
information about growth or individual differences in growth over time as these models
are based on the between-wave covariance matrix and do not explicitly model the mean
structure of the data (Rogosa, 1987, 1988). While AR models are useful when processes
are systematically changing over time, they do not consider individual differences in
growth, nor do they take complete advantage of all the information available. This could
lead to biased model parameters and ultimately to misleading conclusions (Curran,
Harford, & Muthen, 1996).
54
One recent approach that is gaining in attractiveness for examining longitudinal
data is latent growth curve modeling (LGM). LGM allows for modeling of growth and
individual differences over time (McArdle & Epstein, 1987; Meredith & Tisak, 1990).
Techniques from confirmatory factor analysis, repeated measures multivariate analysis of
variance (MANOVA), and structural equation modeling are combined to analyze changes
in a construct longitudinally. LGM is a valuable technique that can allow the researcher
to model constructs such as emerging adulthood. Some methodological concerns that
often pose a problem include missing data and model measurement error; LGM is able to
account for these problems and provide accurate estimates of the results. Clear
advantages exist for using LGM to model substance use over time compared to more
traditional fixed-effects techniques (Curran, Stice, & Chassin, 1997). Given the utility of
these techniques it is surprising that only a paucity of studies exist that employ these
advanced modeling approaches to study substance use over time among teens and
emerging adults (defined by age range), and none examine the emerging adulthood
construct longitudinally.
While LGM is a significant advance from previous statistical techniques it is not
without its faults. One such limitation to LGM is that it assumes that the observed
trajectories of growth represent a sample from a single population of individuals that are
characterized by a single average status and by a single average rate of change. That is to
say that all individuals in the given sample begin at the same place and grow at the same
rate. However, it more likely that a sample comes from more than one population and
that each population has its own unique starting point and growth trajectory. Conversely,
55
latent class growth modeling (LCGM) assumes that the data is comprised of a mixture of
groups, and that the group membership is “latent” or unobserved (B. Muthen, 2001).
Thus, this framework can be used to identify different developmental trends of latent
classes/groups over time. This method is proposed here to allow for the identification of
two or more emerging adulthood trajectory classes.
Longitudinal Studies Using Latent Class Analysis or Growth Mixture modeling
Several studies that employ latent class analysis are reviewed to demonstrate the
utility of such methodology. One study (Jackson et al., 2000) examined conjoint
trajectories of emerging adult substance use over four time points. This study used
growth mixture modeling to identify trajectories of substance use for cigarettes and
alcohol in emerging adults spanning from 18 to 26 years of age. Later, co-morbidity of
initiation and growth was examined by cross-classifying group membership in substance
use categories. Results indicated that five longitudinal types of alcohol or tobacco use
disorders over time exist: non-diagnosing, developmentally limited alcohol use disorder,
chronic alcohol use disorder, chronic tobacco use disorder, and comorbid alcohol and
tobacco use disorder. These various classes of trajectories were discernible on the basis
of family history of alcoholism status and gender. The study also identified etiologically
important third variables such as alcohol expectancies, behavioral under-control, and
childhood stressors. These variables mediated the relationship between family history
and latent class. It appears that parallel changes occurred in substance use during
emerging adulthood, suggesting similar timing of use potentially based on developmental
56
factors. In addition, it appears that individuals might use multiple substances because of
a common vulnerability to each, and not because of directional relations between
substances. The results led to important conclusions that could not have been obtained
using a different methodology; it appears similar developmental timing of use across
substances exists, perhaps indicative of common influences on use. Thus, non-
longitudinal models that do not examine growth over time are insufficient to model
cigarette, alcohol and drug use in youth.
Another study (Anderson, Ramo, Cummins, & Brown, 2010) identified patterns
of alcohol and other drug use, developmental outcomes, and psychosocial variables at
eight time points during emerging adulthood in a sample of alcohol or drug dependent
individuals. Latent class growth modeling was used to identify six latent trajectories
constructed using alcohol and substance use frequency. These classes appeared to be
consistent with developmental transitions. It was also found that initial differences on
education and occupational status were differentially related to substance use trajectories.
More severe drug use trajectories were found to be correlated with high dependence rates,
incarceration, and more treatment. High rates of high school graduation, professional
occupations, marriage/cohabitation, and financial responsibility for children were found
for those with the least alcohol and drug use. Only through longitudinal methodology
would the long-term diversity of substance use outcomes be detectable. Such analysis
permits clarification of the developmental impact of alcohol and drug use in emerging
adults.
57
Finally, a parallel-process growth mixture model was used to examine changes in
alcohol and marijuana use and use of drug and alcohol treatment services over seven
years (Greenbaum & Dedrick, 2007). The sample consisted of 180 adolescents with
serious emotional disturbances. This study used LCGM to identify classes of substance
users to examine rate of change in substance use and found that it was correlated with the
number of alcohol and drug services individuals received. For example, the class
displaying initially low substance use, which subsequently increased sharply in use,
ended up similar to the high initial substance use class over time, and yet received fewer
services. This study underlines the need to employ rate of change and LCGM to study
substance use so that treatment or prevention research can be done effectively.
Together these studies highlight the theoretical and applied utility of employing
longitudinal methodologies to examine how specific processes initiate and change over
time. Furthermore, these techniques allow for the examination of shared variance and
unique contributions of specific mechanisms that could potentially account for substance
use trajectories. While substance use trajectories have been modeled over time using
LGM, latent class analysis and growth mixture modeling have not been used to model
actual developmental trajectories. Namely, behavioral outcomes have largely dominated
the scope of these analyses, while specific developmental processes have not been
examined using these modern methodologies. Specifically, this type of examination has
not yet been conducted using emerging adulthood. The importance of investigating latent
trajectory classes of emerging adulthood over time is important to attain a deeper
theoretical understanding of human development. The identification of latent classes will
58
be beneficial for the prediction of future behavioral and psychosocial outcomes. In
addition, future prevention efforts will be able to use this information to inform program
development based on predicted class membership of participants. Thus, emerging
adulthood is proposed as a potential factor involved in this complex relationship.
Substance Use, Emerging Adulthood and LCGM
Electronic databases (Google Scholar and PubMed) were searched to identify
articles that used latent class analysis to study either emerging adulthood or substance use
during emerging adulthood (accessed August 4, 2011). The databases were searched
using the terms “latent class analysis,” “emerging adulthood,” and “substance use.”
Searches revealed 111 hits. Upon careful examination of the studies it appears that the
studies can be categorized into two distinct groups. The first group, containing the large
majority of the studies, uses LCGM to identify varying trajectories of substance use
behavior during emerging adulthood (Iwamoto, Corbin, & Fromme, 2010; Jackson, Sher,
Rose, & Kaprio, 2009), and the second group uses substance use data during the
emerging adulthood period to study LCGM as a new statistical approach (Lanza, Patrick,
& Maggs, 2010; Tan, Dierker, Rose, & Li, 2011). However, no study as of yet uses
LCGM to identify distinct trajectories of emerging adulthood as a developmental
construct.
A variety of studies have examined substance use trajectories during the transition
to adulthood (Anderson et al., 2010; Auerbach & Collins, 2006; Riggs, Chou, Li, &
59
Pentz, 2007; J Schulenberg et al., 2005; J. S. Tucker et al., 2003; White, Labouvie, &
Papadaratsakis, 2005).
In a comparison of college students and their non-college age peers, White et al.
(2005) found that regardless of college status, higher levels of substance were found
during emerging adulthood use than for peers still in high school. In addition, the non-
college sample differed from the college sample in important ways such that the non-
college sample reported higher levels of cigarette and marijuana smoking during
emerging adulthood.
Another study (J. S. Tucker, Ellickson, Orlando, Martino, & Klein, 2005)
employed growth mixture modeling to identify substance use trajectories from
adolescence to emerging adulthood (age range). As expected, early users were at a
higher risk for poor outcomes (stealing, selling drugs, predatory violence, poor health
outcomes, and poor mental health) at age 23 compared to low-level users and abstainers.
Smoking trajectories from adolescence to emerging adulthood were examined
using discrete mixture modeling by Riggs et al. (2007). Findings indicated that
membership in increasingly high-use trajectories placed individuals at a greater relative
risk for nicotine dependence than lower-use trajectories.
Auerbach and Collins (2006) examined emerging adulthood alcohol use using
latent transition analysis. Results indicated that individuals in lower-level alcohol use
latent classes and participants in moderate- and high-level latent classes were more likely
to stay in their original class over time.
60
Trajectories of marijuana use were examined during the transition to adulthood in
a national sample (Schulenberg et al., 2005). This study found that demographic and
lifestyle characteristics (e.g. ethnicity, educational level, religious importance, and high
school grades) were all related to marijuana use trajectory groups.
Together these studies highlight how trajectories of substance use appear to vary
according to developmental timing. In addition, it is clear that substance use is higher
during emerging adulthood but appears to differ among individuals indicative of the
heterogeneity of this period of life. Non-college attending individuals appear to use more
substances during the emerging adulthood period than college students (White et al.,
2005). Early substance users are at a higher risk for poor outcomes later in life (Riggs et
al., 2007; J. S. Tucker et al., 2005). Class status did not appear to be transitional, such
that members of each substance use class were relatively stable over time (Auerbach &
Collins, 2006). In addition, demographics and lifestyle characteristics were found to be
associated with marijuana trajectory class (Johnston, Bachman, & Schulenberg, 2005).
However, no study to my knowledge has examined trajectories of emerging adulthood as
a developmental construct over time. Thus, the present study is be the first of its kind to
both use LGM and LCGM to identify varying trajectories of emerging adulthood over
time.
Present Study
In the first validation study of the IDEA, Reifman, Arnett and Colwell (2007)
found that when comparing age groups, the IDEA sub-scales for identity exploration,
61
experimentation/possibilities, and negativity/instability (in both samples), and for self-
focus were all highest in the 18 to 23 year old age group, and declined in the older age
groups. Other-focus, exhibited the reverse pattern. The authors also found that across
studies individuals in the emerging adulthood sample scored higher on identity
exploration, other-focus, self-focus, and feeling “in-between” than did younger
respondents (sixth through twelfth graders). While these results are interesting and
indicate a sample in transition, the results need to be extended longitudinally. Without
longitudinal evidence, it is impossible to know if cultural and historical disparities across
age cohorts rather than actual developmental processes account for these differences. It
appears that the IDEA measure captures the essence of emerging adulthood, as indicated
by higher scores, as opposed to other developmental periods. LGM will allow the
identification of the overall trajectory of emerging adulthood in the current sample.
Nevertheless, it is important to investigate not only how individuals experience emerging
adulthood, but also how these experiences differ individually. Not all emerging adults
will experience this period in the same way. In fact, emerging adulthood is characterized
by its great heterogeneity. Thus, LGM is used to examine the overall pattern of
development through emerging adulthood. Subsequently, LCGM is used to examine
developmentally differing trajectories of passing through the emerging adulthood period.
The following hypotheses are proposed:
1. Assess the predictive utility of emerging adulthood status on future demographics
and risk behavior correlates in continuation high school students
62
a. Examine the factorial invariance of the 8-item emerging adulthood
measure over three time points. It is expected that the factor structure will
be invariant over time.
b. Use LGM to examine the overall trajectory shape of emerging adulthood
over three waves of data. Overall, trajectories are expected to exhibit
linear growth.
c. Examine if individuals can be classified into discrete latent groups based
on emerging adulthood status by using LCGM to identify the number of
classes. It is hypothesized that four latent classes will emerge. These
classes will resemble those typically found in latent class substance use
studies (Marti et al., 2010). This is expected because of the consistent
evidence of emerging adulthood and substance use being positively
correlated (Arnett, 2004b, 2005, 2007a; Jackson et al., 2000). Thus the
trajectories should resemble the following: (1) the normative trajectory
class (this class will rise moderately over time as the individuals move
from adolescence to emerging adulthood), (2) early onset class (this class
will enter emerging adulthood earlier than the other trajectory classes and
then decline, thus representing maturation from adolescence to emerging
adulthood, and finally entering adulthood earlier than others), (3) late
onset emerging adult class (this class will rise moderately over time
representing a class of individuals who enter emerging adulthood more
slowly), and (4) the later-onset rapid rising trajectory (this group of
63
individuals will enter emerging adulthood later than the normative or late
onset groups, but will enter it rapidly, and thus have the steepest slope).
d. Regress latent class on measures demographics and risk behaviors at Time
3. Emerging adulthood latent class is expected to predict the measures
from study I in the same direction. More specifically, it is hypothesized
that the latent classes exhibiting the highest emerging adulthood scores at
Time 3 will be related with variables that might indicate emerging
adulthood transitions, such as demographic characteristics and risk
behaviors.
64
Method
Participants and Procedures
Participants are from 24 continuation high schools that were part of a larger drug
prevention study in Southern California (Lisha et al., in press). Schools were recruited as
a convenience sample based on specific criteria such as number of students and ethnic
composition of the school. Typically continuation high school populations attend this
type of school because of extreme truancy, poor grades, drug use, violence, other illegal
behaviors, or other disruptive activities. Continuation high school students have an
increased likelihood of using drugs and alcohol than the general population (Sussman,
2010). The same procedures used in Study I were followed. All procedures were
approved by the university’s Institutional Review Board. See Table 6 for demographic
information.
Survey Measures
Demographics. Participants were asked to report their age, gender, and ethnicity
(coded as African-American, Latino, White, Mixed, and Other) at Time 1. These were
considered as covariates along with highest level of parental education (1 = did not
complete 8
th
grade, 2 = did not complete high school, 3 = completed high school, 4 =
some college or job training, 5 = completed college, and 6 = attended or completed
graduate school).
65
Emerging adulthood was measured using a reduced 8-item scale at all time points.
The scale included at least one item from each subscale. The question stem was “For the
next set of questions think about this time in your life. When we say “this time,” we mean
what is going on now, plus what has gone on in the last few years, plus what you think
your life will be like in the next few years. Think about a 5-year period of time, with right
now in the middle. For each question below, mark the box that best describes this time in
your life. Be sure to put only one check mark per line. Is this period of your life a…” (1 =
definitely not, 2 = probably not, 3 = maybe, and 4 = definitely yes). See Table 2. The
reduced scale was created by conducting an exploratory factor analysis on the 21-item
measure from Time 1. Before rotation, the 21-item scale revealed only one factor, three
factors had eigenvalues above one, but eigenvalues dropped significantly after the first
factor (eigenvalue for factor 1 = 8.6, factor 2 = 1.5, and factor 3 = 1.1). A promax
(oblique) rotation revealed three factors (eigenvalue for factor 1 = 5.1, factor 2 = 3.8, and
factor 3 = 2.4). The shortened 8-item scale (alpha = .88) was created by selecting the 8
items that loaded the most on the first factor. See Table 5.
Table 5
Reduced Emerging Adulthood Measure
Item Subscale
1. time of independence? Self-Focused
2. time of open choices? Identity Exploration
3. time of defining yourself? Identity Exploration
4. time of planning for the future? Identity Exploration
5. time of seeking a sense of meaning? Identity Exploration
6. time of deciding on your own beliefs and values? Identity Exploration
7. time of learning to think for yourself? Identity Exploration
8. time of gradually becoming an adult? Feeling "In-Between"
66
Time 3 Measures
Demographics and measures regressed on the IDEA and its subscales. Various
categories of correlates were selected to look at their different relations with the IDEA
factors. Certain demographic variables (i.e. job status, marital status, parenthood) were
selected as they might provide concrete markers of the transition to young adulthood.
Control variables included age (in years), gender, and ethnicity (indicator coded as
Latino/Hispanic, African American/ Black, White/Caucasian, mixed, or other).
Participants were also asked their job status (0 = no job, 1 = have a job), whether they
were married (0 = no, 1 = yes), and whether they were a parent (0 = no, 1 = yes).
Risk Behaviors. Risk behaviors may indicate difficulties with resolving the
emerging adulthood process. Participants were asked, “How many times in the last
month have you used…” each of 12 substance use categories (cigarette, alcohol use,
getting drunk on alcohol, marijuana, cocaine, hallucinogens, stimulants, inhalants,
ecstasy, pain killers, tranquilizers, or other hard drugs). The responses to the last eight
drug categories (cocaine through other hard drugs) were summed to form a hard drug
index. All the substance use items were measured on 12-point scales (0 times to over 100
times). The present study utilized four categories of drug use: drunk on alcohol,
cigarettes, marijuana and hard drug use. The reliability of the drug-use item format used
here has been previously established (Graham et al., 1984). In addition, participants were
asked, “During the past 30 days, with how many people have you had sexual
intercourse?” (0 = 0 people, 1 = 1 person, 2 = 2 people, 3 = 3 people, 4 = 4 people, 5 = 5
people, 6 = 6 people, 7 = 7 people, 8 = 8 people, 9 = 9 people, 10 = 10 people, 11 = more
67
than 10 people), and “Was a condom used the last time you had sexual intercourse?” (0 =
no, 1 = yes).
Substance abuse (SADS). Substance abuse was measured using four items. One
sample items includes, “In the last 12 months, have you kept using alcohol or drugs even
though it was keeping you from meeting your responsibilities at school or home?” (1 =
yes, 2 = no). Items measured four categories of abuse: (1) recurrent substance use
resulting in failure to fulfill major role obligations at work, school, or home, (2) recurrent
substance use in situations in which it is physically hazardous, (3) recurrent substance-
related legal problems, and (4) continued substance use despite social or interpersonal
problems caused or exacerbated by use (Dennis, Titus, White, Unsicker, & Hodgkins,
2002). The Cronbach’s alpha was 0.72.
Drug problems/consequences (PEI). PEI was measured using an 11-item scale.
Items included, “In the last 12 months, how many times have you taken or sold things
that weren’t yours to get or pay for alcohol or other drugs?” and “In the last 12 months,
how many times have you done personal favors for people to get or pay for alcohol or
other drugs?” (1 = none, 2 = once or twice, 3 = sometimes (3 to 9 times), 4 = often (10 or
more times)) (Sussman et al., 2000; Winters, 1990). Cronbach’s alpha for this scale was
0.82.
Intercept and slope. Two latent variables (intercept and slope) will be created for
emerging adulthood. They will be created using emerging adulthood indicators at all
three waves of data. This is a common methodology used for latent growth curve
modeling (Curran et al., 1997; Fleming, Mason, Mazza, Abbott, & Catalano, 2008).
68
Statistical Analyses
Test factorial invariance across Time Points
The factorial invariance of the 8-item reduced IDEA was examined to determine
if the three-factor structure of the IDEA is consistent across time points. Structural
equation modeling (SEM) allows for the factor structures to be modeled and compared
simultaneously across all time points. Specifically, CFA allows for the researcher to test
the hypothesized factor structure of the data while constraining parameters to be equal
across two or more groups (Deng, Doll, Hendrickson, & Scazzero, 2005). The indicator
loadings can then be compared between groups and if these loading are not significantly
different then factorial invariance will be assumed (Drasgow, 1984; Drasgow & Kanfer,
1985). Model fit indicators will also be used to assure good model fit (Chi-square,
degrees of freedom, p, CFI, and RMSEA with 95% confidence intervals). In the case that
full factorial invariance is not achieved, partial invariance can be considered satisfactory
(Byrne, Shavelson, & Muthén, 1989). Factorial invariance provides evidence that we are
measuring that which we claim to be measuring.
Conceptual models of latent growth curve modeling
Using Singer’s (1998) influential article the basic form of the overall model can
be explained as follows:
Unconditional linear growth model. A series of repeated measurements, Yit (here
three repeated measures, Yi1, Yi2, Yi3), where i represents an individual at time t are used
69
to form the individual linear growth model. This model can be referred to as the Level-1
within-person model; such that:
Yit = αi + βiλi + εi,
where αi represents the intercept of an individual’s latent growth trajectory (i.e. the
average initial level of the construct measured at Time 1), βi represents the latent slope
for an individual’s growth trajectory (i.e. as in regression, the average unit change in the
outcome, Yi, between two consecutive measurements), λi represents the sequential time
points at which the measurement was taken, and lastly, εi represents the residual variance
for an individual (i.e. error).
The Level-2, or between-person model can be examined such that:
αi = µα + ζαi, and βi = µβ + ζβi,
The latent intercept factor (αi) and the latent slope factor (βi) are random variables.
These latent factors can be represented by the group mean intercept (µ α), the group mean
slope (µβ), and individual variation for both (ζαi and ζβi). The Level-2 model ass umes
that the latent individual growth trajectory factors (αi and βi) are a function of group
mean latent growth factors (µα and µβ) and individual variation (ζαi and ζβi). In
addition, this model assumes that there are no other predictors that are responsible for the
variation of the individual growth trajectory factors. The Level-2 model is known as the
unconditional growth model (Curran, 2000) because the individual growth trajectories are
not a function of outside predictors. Figure 5 depicts this model.
Figure 5. Basic two-factor, three time point latent
factor, three time point latent growth model.
70
71
Using this methodology, individual differences in initial levels and growth over
time can be examined by looking at variances of the intercepts and slopes (Cheong et al.,
2003). One way to consider LGM is within a confirmatory factor analysis framework
(Curran et al., 1997). Using this framework, the first factor can be defined as the starting
point of growth (intercept: factor loadings are set to 1.0 at all time points), and the second
factor defines the slope, rate, or trajectory of change. For the slope, the first factor
loading (T1) will be set to 0.0 to indicate where the growth trajectory begins.
Analyses will be completed using Mplus software (L. Muthen & Muthen, 2007).
Growth will first be modeled as linear (slope values specified as 0, 1, 2). Model-fit
statistics were evaluated to make appropriate modifications.
Goodness of fit indices were evaluated by looking at overall model fit (absolute,
parsimony, and comparative fit). Absolute fit was evaluated using the model chi-square
(χ
2
), where a non-significant p-value indicates good model fit, and a significant p-value
indicates poor model fit. Parsimony was indicated by the root mean square error of the
approximation (RMSEA) where values below 0.06 indicated good fit (Brown, 1989).
Comparative fit was assessed using the Tucker-Lewis index (TLI) and the Comparative
Fit Index (CFI) where values above 0.90 indicate adequate fit. Because of the large
sample size it was recommended to not rely on the chi-square alone as it can increase the
probability of Type 1 error might be increased (Cheung & Rensvold, 2000). Structural
path significance will be evaluated using z-scores (values ± 1.96 or greater indicated
significance at α = 0.05, two-tailed).
72
Latent class analysis and growth mixture modeling
Growth mixture modeling allows both categorical and continuous latent variables
to be combined into the same model. A mixture of latent growth trajectory classes will
be hypothesized and tested. Latent class growth mixture modeling is a statistical method
that provides trajectory classification probabilities for each participant, classifies
individuals in their most likely class based on the previous probabilities, and allows for
regression of class membership to be made on covariates (B. Muthén, 2001; B. Muthén &
Shedden, 1999). The model is also capable of estimating class membership probability
as a function of covariates and incorporating outcomes of the latent class variable. That
is to say that for each class, the values of the latent growth parameters can be influenced
by the chosen covariates in the model.
The Figure 5 model is a graphical representation containing two continuous latent
growth curve variables, η
1
and η
2
(η
1
= emerging adulthood intercept and η
2
= emerging
adulthood slope), and a latent categorical variable, C, with K classes C
i
= (c1, c2, …, c
K
)’,
where c
i
= 1 if the individual i belongs to class k and zero otherwise. Circles represent all
the latent variables. Following previous research (Willett & Sayer, 1994), the latent
continuous growth variable portion of the model is represented by a multivariate growth
model with multiple measures (emerging adulthood) with multiple (three) time point
indicators each, Y. A categorical latent variable represents the latent trajectory classes
that underlie the latent growth variables, η.
Thus, first, the optimal number of growth trajectories was ascertained based on
select fit indices and theoretical models. There is not one established fit index that should
73
be used. Bayesian information criterion (BIC) which indicates a more parsimonious
model when the value is low (high log likelihood estimate and low number of
parameters) (B. Muthén, 2002; Schwarz, 1978) and differences of 10 or more considered
as evidence favoring one model over another (Raftery, 1995). Each individual is
classified into their most likely class, and thus a table is created where rows correspond to
individuals and columns correspond to conditional probabilities of class belongingness
(Nagin, 1999). In addition, an entropy summary statistic indicates the quality of the
classification. Values of this statistic range from 0 to 1; values closer to 1 represent better
classification quality (B. Muthen et al., 2002). The Lo-Mendell-Rubin likelihood ratio
test of model fit is used to quantify the likelihood that the data can be described by a
model with one-less class and a p-value smaller than 0.05 indicating that the additional
class significantly improves fit over a model with k - 1 classes (Lo, Mendell, & Rubin,
2001). For the decision on the number of classes in the final model, the Lo-Mendell-
Rubin test was established by the more precise and accurate computationally demanding
parametric bootstrapped likelihood ratio test with 100 draws (McLachlan & Peel, 2000;
Nylund, Asparouhov, & Muthén, 2007).
74
Figure 6. Graphical representation of the associative growth mixture model without
covariates.
75
Incorporation of Covariates
Next, a set of covariates (age, gender, condition) were included as covariates.
These covariates are used as predictors of both latent continuous and latent class
variables, since this model allows proportions to be predicted from prior information
and/or subject-specific variables. Evidence indicates that class membership is better
defined when covariates are incorporated into the model (B. Muthén, 2004). By
including covariates, the growth mixture model takes into account relationships occurring
between covariates and class membership η
ik
(personal growth parameters within each
class k). Here, there are q = 3 covariates measured (x
age
, x
gender
, x
condition
). These
covariates are included as they might influence emerging adulthood as measured such
that:
η
ik
= Α
k
+ Γ
k
x
ik
+ ζ
k
where η
ik
is a vector of growth parameters of person i in latent class k, Α
k
is a (m x 1)
matrix for the observed variables y within the k
th
class, Γ
k
is a matrix of within class
regression coefficients of growth parameters on q covariates, and ζ
k
is a m-dimensional
residual vector of growth parameters for i in latent class k. The ζ
k
is normally distributed
and the class specific variance-covariance matrix, ψ
k,
is 0. Covariates are important to
consider in growth mixture modeling. Figure 6 depicts the latent growth model with
covariates.
76
Figure 7. Graphical representation of the associative growth mixture model with
covariates.
Regression with Time 3 Outcomes
Subsequently, participants were assigned to the trajectory class for which they had
the highest probability of membership. Logistic and least squares regression analysis
were used to model each of the 11 outcomes as a function of trajectory class membership.
It should be noted that both TND intervention and control participants were included in
all analyses to use the maximum sample size possible. The intervention was
implemented after Time 1, and its effect on substance use will likely be minimal at Time
77
3. Nonetheless, the analyses controlled for the TND group membership (intervention vs.
control), as well as certain demographic characteristics. Using these models as a basis, it
was possible to evaluate the association between class membership and each outcome by
looking at the significance of the linear contrasts for each of the pairwise class
comparisons. This portion of the analysis will be completed using SAS (SAS, 2008).
Missing Data
It is important to properly account for missing data in the sample as not
accounting for it can significantly bias results and subsequently invalidate the findings
from the research. Part of the difficulty in dealing with missing data is that data can be
missing for more than one reason. Researchers can create intentional missing patterns or
missing data can be caused in an unanticipated fashion by such things as respondent
error, accidental or purposeful skipping, absenteeism or lack of time to complete the
survey. As the present study plans to use data over three waves of data collection, it is
recognized that missing data is important and must be addressed accordingly to assure
reliable model estimates.
Previously, listwise deletion, imputation, multiple imputation or pairwise deletion
procedures have been used frequently to deal with missing data. However, these methods
often result in a loss of power and biased estimates of means, variances, and covariances
unless the data is truly missing completely at random (MCAR), which is a rare case in
social science research (Scheffer, 2002). The structural equation modeling analyses
present in this paper will use the full information maximum likelihood (FIML) method in
78
Mplus to account for any missing data (Little & Rubin, 1987; Rubin, 1976). Principles
for using ML and likelihood-based procedures on missing/incomplete data is first
described by Rubin (1976), and a review can be found by Little and Rubin (1987). The
FIML method has been shown to produce more accurate estimates in model estimations
by adjusting for the uncertainty caused by missing data (Collins, Schafer, & Kam, 2001;
McArdle & Hamagami, 1992).
FIML directly estimates model parameters and standard errors using all the raw
data that is available across all the waves of data collected rather than imputing or filling
in missing values. In a wide variety of missing data situations, this type of estimation has
been shown to be the optimal manner to deal with missing data to date (Collins et al.,
2001).
Dealing with multi-level data
Specific multi-level data procedures are required when using complex data
collection designs (i.e. stratification and clustering) to properly estimate parameters and
error terms (McArdle & Hamagami, 1992; Siddiqui, Hedeker, Flay, & Hu, 1996; Singer,
1998). The reason these particular procedures are necessary is that complex research
designs do not intrinsically account for probable dependence among the units of analysis
in the statistical estimations. Oftentimes the problem arises when students (unit of
analysis) from the same school (unit of selection) are more likely to be similar to each
other than are students from other schools or groups on certain fundamental
characteristics such as attitudes and behaviors. If this is the case then the assumption of
79
independence can no longer be considered to be true and multi-level modeling procedures
are required to account for this violation. If the fact that students are nested within
schools is not accounted for then standard errors can be underestimated, Type I errors
increase, and this could result in invalid conclusions (Murray, Varnell, & Blitstein, 2004).
Thus, intraclass correlations cannot be ignored (Murray & Hannan, 1990). The
program condition (treatment or control) will be considered a fixed effect, while school
(which was the unit of randomization) will be considered a random effect. This allows
for both the statistical accounting of intraclass correlation within clustered units (school)
on computed significance levels, and for generalization of the findings to a larger sample.
Additional covariates included age, ethnicity, gender, and baseline score on given
correlate.
Results
Descriptive statistics of emerging adulthood items across time
Descriptive statistics of each item over time are presented in Table 5, including
mean, standard deviation, skew and kurtosis. Table 6 presents correlations among items
over time.
80
Table 6
Descriptive statistics of IDEA items across times
Mean SD Skewness Kurtosis
T1 - Item 1 (n = 1,603) 3.41 0.83 -1.35 1.04
T1 - Item 2 (n = 1,602) 3.47 0.74 -1.44 1.79
T1 - Item 3 (n = 1,596) 3.38 0.77 -1.21 1.13
T1 - Item 4 (n = 1,597) 3.65 0.66 -2.14 4.68
T1 - Item 5 (n = 1,592) 3.38 0.75 -1.14 0.98
T1 - Item 6 (n = 1,596) 3.45 0.77 -1.34 1.20
T1 - Item 7 (n = 1,597) 3.64 0.68 -2.10 4.43
T1 - Item 8 (n = 1,594) 3.59 0.68 -1.84 3.34
T2 - Item 1 (n = 1,173) 3.56 0.78 -1.82 2.56
T2 - Item 2 (n = 1,176) 3.63 0.70 -2.09 4.16
T2 - Item 3 (n = 1,175) 3.46 0.81 -1.47 1.42
T2 - Item 4 (n = 1,171) 3.77 0.61 -3.08 9.61
T2 - Item 5 (n = 1,172) 3.48 0.80 -1.55 1.74
T2 - Item 6 (n = 1,174) 3.61 0.76 -2.11 3.85
T2 - Item 7 (n = 1,171) 3.73 0.65 -2.80 7.77
T2 - Item 8 (n = 1,170) 3.76 0.60 -2.92 8.88
T3 - Item 1 (n = 639) 3.56 0.83 -1.86 2.43
T3 - Item 2 (n = 639) 3.62 0.75 -2.17 4.16
T3 - Item 3 (n = 639) 3.59 0.82 -2.03 3.16
T3 - Item 4 (n = 639) 3.86 0.51 -4.24 18.59
T3 - Item 5 (n = 639) 3.64 0.76 -2.25 4.36
T3 - Item 6 (n = 639) 3.71 0.72 -2.68 6.42
T3 - Item 7 (n = 638) 3.85 0.57 -4.08 16.08
T3 - Item 8 (n = 638) 3.88 0.50 -4.54 21.02
Note: T1 = Time 1, T2 = Time 2, T3 = Time 3.
81
Table 7
Correlation matrix among items across time
Time 1 Time 2 Time 3
Item 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8
T1 - Item 1 1.00
T1 - Item 2
0.58 1.00
T1 - Item 3 0.50 0.51 1.00
T1 - Item 4
0.38 0.43 0.49 1.00
T1 - Item 5
0.44 0.46 0.56 0.53 1.00
T1 - Item 6 0.43 0.47 0.56 0.51 0.61 1.00
T1 - Item 7
0.43 0.48 0.53 0.52 0.53 0.58 1.00
T1 - Item 8 0.40 0.43 0.45 0.55 0.47 0.49 0.57 1.00
T2 - Item 1 0.24 0.17 0.18 0.13 0.16 0.17 0.17 0.10 1.00
T2 - Item 2 0.13 0.18 0.14 0.15 0.19 0.17 0.15 0.14 0.47 1.00
T2 - Item 3 0.15 0.14 0.23 0.11 0.18 0.21 0.17 0.11 0.40 0.45 1.00
T2 - Item 4
0.08 0.15 0.09 0.18 0.13 0.18 0.14 0.11 0.39 0.41 0.44 1.00
T2 - Item 5
0.12 0.16 0.18 0.16 0.18 0.18 0.14 0.13 0.36 0.41 0.49 0.54 1.00
T2 - Item 6
0.10 0.16 0.13 0.15 0.16 0.21 0.15 0.12 0.30 0.38 0.41 0.48 0.56 1.00
T2 - Item 7 0.09 0.19 0.15 0.18 0.16 0.18 0.19 0.15 0.40 0.43 0.42 0.51 0.49 0.54 1.00
T2 - Item 8 0.09 0.11 0.11 0.14 0.12 0.15 0.12 0.17 0.38 0.43 0.36 0.51 0.42 0.44 0.49 1.00
T3 - Item 1 0.14 0.11 0.10 0.13 0.13 0.12 0.15 0.06 0.28 0.22 0.16 0.11 0.15 0.10 0.11 0.11 1.00
T3 - Item 2 0.10 0.12 0.14 0.06 0.05 0.08 0.07 0.07 0.11 0.15 0.12 0.07 0.13 0.08 0.06 0.10 0.27 1.00
T3 - Item 3
0.11 0.15 0.16 0.15 0.19 0.16 0.16 0.16 0.20 0.12 0.18 0.10 0.14 0.17 0.14 0.09 0.31 0.24 1.00
T3 - Item 4 0.06 0.10 0.08 0.05 0.05 0.08 0.05 0.00 0.09 0.05 0.12 0.08 0.12 0.07 0.04 0.09 0.15 0.22 0.37 1.00
T3 - Item 5
0.06 0.13 0.13 0.10 0.16 0.14 0.17 0.09 0.11 0.16 0.15 0.13 0.25 0.23 0.15 0.16 0.22 0.21 0.48 0.40 1.00
T3 - Item 6 0.09 0.13 0.13 0.10 0.07 0.12 0.17 0.11 0.11 0.15 0.14 0.12 0.20 0.23 0.10 0.12 0.22 0.21 0.38 0.35 0.55 1.00
T3 - Item 7 0.12 0.07 0.06 0.09 0.07 0.05 0.09 0.02 0.06 0.04 0.10 -0.02 0.06 0.11 0.10 0.07 0.18 0.24 0.33 0.40 0.31 0.38 1.00
T3 - Item 8 0.05 0.06 0.06 0.11 0.04 0.06 0.09 0.13 0.08 0.11 0.15 0.05 0.16 0.05 0.08 0.14 0.18 0.16 0.28 0.34 0.26 0.20 0.33 1.00
Note: T1 = Time 1, T2 = Time 2, T3 = Time 3.
82
Test of Factorial Invariance
Longitudinal invariance analysis was conducted in several steps. Table 8 presents
the results of model fits and comparisons.
The initial 8-item baseline model was tested for configural invariance. This
model was not acceptable (CFI = 0.932, TLI = 0.924, RMSEA = 0.038). Items with the
lowest factor loadings across time points were eliminated one by one until configural
invariance was achieved. This was achieved with a 5-item structure (CFI = 0.975, TLI =
0.969, RMSEA = 0.031). The items that were eliminated included “Time of
independence?” “Time of planning for the future?” and “Time of gradually becoming an
adult?”
Factor loadings were then constrained to be equal across time to test for weak
invariance. The weak invariance model was acceptable as it satisfied the fit indices well
(CFI = 0.966, TLI = 0.962, RMSEA = 0.035). The fact that metric invariance (“weak
Table 8
Model fit of various invariance models
Model df
χ
2
∆χ
2
CFI TLI RMSEA RMSEA
Lower CI
RMSEA
Upper CI
Configural invariance 249 948.41 - 0.932 0.924 0.041 0.038 0.044
Configural invariance
a
87 228.68 - 0.975 0.969 0.031 0.026 0.036
Weak invariance 95 286.85 58.18* 0.966 0.962 0.035 0.030 0.040
Strong invariance 105 399.59 112.74* 0.947 0.947 0.041 0.037 0.045
Strict invariance 115 516.97 117.38* 0.928 0.934 0.046 0.042 0.050
*p < .0001
Note:
a
Model fit was achieved by eliminating 3 items. All subsequent models are based on the 5 remaining items.
83
invariance”) held, indicates that the items were related to the latent factor equivalently
across time, or more simply, that the same latent factor was being measured at each time.
Further, equality of intercepts was imposed on the model to test for strong invariance.
The strong invariance model also had acceptable values on the fit indices (CFI = 0.947,
TLI = 0.947, RMSEA = 0.041). Finally, equality of item uniqueness across time was
further imposed to test for strict invariance. The strict invariance model was also
satisfied with the fit indices (CFI = 0.928, TLI = 0.934, RMSEA = 0.046). Thus, factorial
invariance was achieved using the 5-item model and further longitudinal analyses are
acceptable.
A score was computed for each individual based on the factor loadings of the
strong invariance model. This score was used in subsequent growth curve models and
other analyses.
Longitudinal Growth Curve Model
First, a latent growth model was used to study the growth of emerging adulthood
over time. Emerging adulthood was modeled across three time points (McArdle &
Epstein, 1987; Meredith & Tisak, 1990). LGM fits under the structural equation
modeling framework and uses unobserved latent factors to examine change over time
(Stoolmiller, Duncan, Bank, & Patterson, 1993; Willett & Sayer, 1994). The basic
growth model is composed of two unobserved latent factors (intercept and slope) and
repeated measures of each construct at each time point (see Figure 5). For purposes of
comprehension, one might view this model as a confirmatory factor analytic model. All
84
the factor loadings of the repeated measures and the intercept of the growth curve, the
first latent factor, were fixed at 1.0 to indicate the start of the growth curve at the first
time point. The second latent factor, the slope, represents the rate of change of the
trajectory of the repeated measures over time. The linear factor loadings are fixed at 0 to
2 with equal time between each measure (0, 1, 2). Means were generated to represent the
latent intercept (starting point) and latent slope over all participants. Variances were
estimated and represent the variance of each individual over the overall growth
parameters.
A two-factor LGM was estimated for three repeated measures of emerging
adulthood over time. This model was found to fit the data well, χ
2
(1) = 1.06, p = 0.30,
TLI = .999 (Tucker-Lewis fit index; TLI) (L. R. Tucker & Lewis, 1973), CFI = .999
(comparative fit index; CFI) (Bentler, 1990), and RMSEA = .006 (root mean square error
of the approximation, RMSEA) (Browne & Cudeck, 1993). The slope factor (µ = 0.121,
p < .0001) and the intercept factor (µ = 3.428, p < .0001) were significantly different
from zero, indicating that overall individuals did not start at a score of zero in emerging
adulthood and that there was significant growth in emerging adulthood over time. The
factor loadings were equally spaced (0, 1, 2) to reflect the hypothesized linear increase.
The variance component for the slope (ψ = 0.024, p = .073) was not significant, but the
variance component for the intercept (ψ = 0.127, p < .0001) factor was statistically
significant; indicating that there were significant individual differences in the initial
levels of emerging adulthood over time. The significant trend for the slope factor
variance indicates possible individual differences in rate of change of emerging
85
adulthood over time. The intercept and slope were not significantly correlated (r = -.024,
p = .13), signaling no relationship between initial status of emerging adulthood and
changes over time.
Inspection of the observed and estimated means (see Figure 6) indicates that there
is substantial variation in trajectories. The estimated mean graph appears to be too
simple to truly represent the observed trajectories.
Figure 8. Observed latent mean trajectories without covariates (N = 100).
86
Figure 9. Estimated mean trajectories without covariates (N = 100).
Thus, we had sufficient reason to proceed with the LCGM; however, first we
estimated the identical growth model with the addition of covariates (condition, age,
gender). This model was found to fit the data well, χ
2
(4) = 8.73, p = 0.06, TLI = .979
(Tucker-Lewis fit index; TLI) (L. R. Tucker & Lewis, 1973), CFI = .936 (comparative fit
index; CFI) (Bentler, 1990), and RMSEA = .027 (root mean square error of the
approximation, RMSEA) (Browne & Cudeck, 1993). The slope factor (µ = 0.669, p <
.001) and the intercept factor (µ = 1.850, p < .0001) were significantly different from
zero, indicating that overall individuals did not start at a score of zero in emerging
adulthood and that there was significant growth in emerging adulthood over time. The
factor loadings were equally spaced (0, 1, 2) to reflect the hypothesized linear increase.
The variance component for the slope (ψ = 0.022, p = .105) was not significant, but the
variance component for the intercept (ψ = 0.111, p < .0001) factor was statistically
87
significant; indicating there were significant individual differences in the initial levels of
emerging adulthood over time. The significant trend for the slope factor variance
indicates possible individual differences in rate of change of emerging adulthood over
time. The intercept and slope were not significantly correlated (r = -.018, p = .23),
signaling no relationship between initial status of emerging adulthood and changes over
time.
Figure 10. Overall sample and estimated mean.
88
Figure 11. Observed mean trajectories with covariates (N = 100).
Figure 12. Estimated mean trajectories with covariates (N = 100).
89
Due to the substantial variance observed in the estimated and observed mean
trajectories and the variances it was determined to proceed with the planned latent class
growth modeling analyses.
Mixture Modeling: Extracting Trajectories
Models which had between 1 and 7 classes were fit to the emerging adulthood
data. The extraction process was performed twice: once without covariates and once with
covariates (program, age, sex). In the first model run without covariates (see Table 8),
based on careful examination of the entropy, BIC, AIC, Adjusted BIC, the Vuong-Lo-
Mendell-Rubin Likelihood ratio test, and the Bootstrapped parametric likelihood ratio
test it was determined that the optimal solution was the 3-class solution (entropy = 0.85,
BIC = 5357.23, AIC = 5281.47, Adjusted BIC = 5312.76, Vuong-Lo-Mendell-Rubin
Likelihood p-value = 0.00, Bootstrapped parametric p-value = 0.00). Significant model
fit improvements were found up to three classes. Class 2 was the largest (85.3%),
followed by Class 1 with 8.5% of the sample, and Class 3 with 6.2% of the sample based
on final class counts and proportions for the latent class patterns based on the estimated
model.
90
In the second extraction process with the addition of the covariates, the number of
classes extracted was identical (see Table 10).
Based on both the present results and other evidence (B. Muthén, 2001; L.
Muthén & Muthén, 2007), it was determined to use the solution with the covariates in the
model. Significant model fit improvements were found up to three classes. Thus it was
concluded that the optimal solution was the three class solution (entropy = 0.85, BIC =
Table 9
Model Fit Criteria for 1 to 7 Class Models in LCGA without covariates
Class Entropy AIC BIC Adjusted BIC Number of
Free
Parameters
Vuong-Lo-
Mendell-Rubin p-
value
Bootstrapped
parametric
likelihood ratio
test
C=1 / 5994.29 6037.58 6012.17 8 / /
C=2 0.94 5551.70 5611.22 5576.28 11 0.00 0.00
C=3 0.85 5281.47 5357.23 5312.76 14 0.00 0.00
C=4 0.78 5151.65 5243.65 5189.64 17 0.11 0.00
C=5 0.74 5032.83 5141.06 5077.53 20 0.05 0.00
C=6 0.75 4962.17 5086.64 5013.57 23 0.57 0.00
C=7 0.75 4898.14 5038.84 4956.24 26 0.28 0.00
Table 10
Model Fit Criteria for 1 to 7 Class Models in LCGA with covariates (program, age, gender)
Class Entropy AIC BIC Adjusted BIC Number of
Free
Parameters
Vuong-Lo-
Mendell-Rubin p-
value
Bootstrapped
parametric
likelihood ratio
test
C=1 / 5929.86 6005.59 5961.11 14 / /
C=2 0.94 5489.70 5597.89 5534.35 20 0.00 0.00
C=3 0.85 5220.42 5361.06 5278.46 26 0.00 0.00
C=4 0.78 5088.35 5261.45 5159.79 32 0.14 0.00
C=5 0.74 4976.54 5182.09 5061.37 38 0.04 0.00
C=6 0.76 4901.88 5139.89 5000.11 44 0.83 0.00
C=7 0.77 4870.10 5140.55 4981.71 50 0.24 0.00
91
5361.06, AIC = 5220.42, Adjusted BIC = 5278.46, Vuong-Lo-Mendell-Rubin Likelihood
p-value = 0.00, Bootstrapped parametric p-value = 0.00). Class 2 was the largest (85%),
followed by Class 1 with 8.7% of the sample, and Class 3 with 6.3% of the sample based
on final class counts and proportions for the latent class patterns based on the estimated
model. Because the solutions did not change much with the addition of covariates and
the estimates remained stable it was determined to use the solution based on the addition
of the covariates.
Figure 13. Final multiple group model of emerging adulthood trajectories over time
adjusting for covariates.
92
Figure 14. Observed individual values with estimated latent class means for Class 1.
Figure 15. Fitted individual values with estimated latent class means for Class 1.
93
Figure 16. Observed individual values with estimated latent class means for Class 2.
Figure 17. Fitted individual values with estimated latent class means for Class 2.
94
Figure 18. Observed individual values with estimated latent class means for Class 3.
Figure 19. Fitted individual values with estimated latent class means for Class 3.
95
Description of Trajectory Classes
Class 1 (8.7%) represented a group of subjects who reported a decrease in
emerging adulthood status over time. While subjects in Class 1 started close to the
highest in emerging adulthood, they ended the lowest of the three classes. (Browne &
Cudeck, 1993). The slope factor (µ = -0.364, p < .05) and the intercept factor (µ = 2.454,
p < .0001) were significantly different from zero, indicating that overall the individuals
did not start at a score of zero in emerging adulthood and that there was significant
decline in emerging adulthood over time. The intercept and slope were not significantly
correlated (r = .02, p = .06), signaling no relationship between initial status of emerging
adulthood and changes over time.
The second class, the largest class (85.1%), represented a group of subjects who
consistently reported stable levels of emerging adulthood. The slope factor (µ = 0.544, p
= 0.135) was not significantly different than zero, and the intercept factor (µ = 2.727, p <
.0001) was significantly different from zero, indicating that overall the individuals did not
start at a score of zero in emerging adulthood, but that there was not significant growth in
emerging adulthood over time. The intercept and slope were not significantly correlated
(r = .02, p = .06), signaling no relationship between initial status of emerging adulthood
and changes over time. The emerging adulthood level for this group was the highest
across all time points compared to the other two classes.
For Class 3 (6.3%), the smallest class, the subjects reported a steady increase in
emerging adulthood over the three survey waves. This class began at the lowest level of
96
emerging adulthood at the first wave, but rose steadily. The slope factor (µ = 0.996, p <
.0001) and the intercept factor (µ = 0.975, p < .0001) were significantly different from
zero, indicating that overall the individuals did not start at a score of zero in emerging
adulthood and that there was significant growth in emerging adulthood over time. The
intercept and slope were not significantly correlated (r = .02, p = .06), signaling no
relationship between initial status of emerging adulthood and changes over time.
Intercept and slope means were compared across groups. It was revealed that all
the groups differed significantly from each other for both intercept and slope.
Specifically, the intercept for Class 2 (M = 3.54, SD = 0.11) was the highest, followed by
Class 1 (M = 3.26, SD = 0.16) and Class 3 (M = 1.91, SD = 0.21). The intercept for Class
3 was considerably lower than the other classes. The groups also differed in their slopes;
the slope for Class 3 (M = 0.78, SD = 0.14) was the steepest and was positive (linear
increase in emerging adulthood over time), followed by Class 1 (M = -0.40, SD = 0.13)
which exhibited a decrease in emerging adulthood over time. Lastly Class 2 had the
smallest slope factor (M = 0.11, SD = 0.07) which differed significantly from the other
groups but was not significant in itself, thus staying at a relatively stable level of
emerging adulthood over time.
97
Table 11. Association between latent class membership and baseline demographics.
Class 1 Class 2 Class 3
87 (5.27%) 1467 (88.86%) 97 (5.88%)
Age, continuous, mean (SD) 16.95 (0.10)
a
16.80 (0.02)
a
16.39 (0.09)
b
Gender
Female 31 (35.6) 639 (43.6) 26 (26.80)
Male 56 (64.4)
ab
828 (56.4)
a
71 (73.20)
b
Ethnicity
African-American 2 (2.4) 76 (5.3) 3 (3.3)
Latino 66 (77.6) 902 (63.1) 64 (70.3)
Mixed 5 (5.9) 200 (14.0) 10 (11.0)
Other 2 (2.4) 91 (6.4) 8 (8.8)
White 10 (11.7)
a
160 (11.2)
a
6 (6.6)
a
Parent
Yes and live with child(ren) 2 (2.3) 69 (4.7) 5 (5.2)
Yes and do not live with child(ren) 0 (0.0) 22 (1.5) 3 (3.1)
No 84 (97.7)
a
1364 (93.8)
a
89 (91.7)
a
Highest education between parents
Did not complete 8th grade 8 (9.6) 132 (9.4) 10 (11.3)
Did not complete 12th grade 30 (36.1) 330 (23.6) 23 (28.4)
Completed high school 22 (26.5) 410 (29.3) 26 (29.6)
Some college or job training 10 (12.2) 308 (22.0) 6 (6.8)
Completed college (4 years) 9 (10.8) 169 (12.1) 14 (15.9)
Attended or completed graduate schoool 4 (4.8)
a
52 (3.6)
a
7 (8.0)
a
Job
No 72 (84.7) 1159 (79.7) 81 (83.5)
Yes 13 (15.3)
a
296 (20.3)
a
16 (16.5)
a
30-day cigarette use, continuous, mean (SD) 2.20 (2.62)
a
2.47 (2.87)
a
2.86 (3.43)
a
30-day alcohol use, continuous, mean (SD) 2.38 (2.10)
a
2.03 (1.60)
a
2.07 (2.03)
a
30-day drunk on alcohol, continuous, mean (SD) 2.12 (2.14)
a
1.74 (1.41)
b
1.67 (1.41)
b
30-day marijuana use, continuous, mean (SD) 3.10 (3.38)
b
2.74 (3.04)
a
3.61 (3.84)
b
PEI, continuous, mean (SD) 1.30 (0.44)
ab
1.26 (0.38)
a
1.35 (0.54)
b
SADS, continuous, mean (SD) 1.75 (0.33)
a
1.75 (0.30)
a
1.72 (0.34)
a
30-day number of sexual partners, continuous,
mean (SD)
1.78 (2.37)
a
2.01 (2.38)
a
2.56 (3.10)
b
a, b, ab
The equality of means for each variable was examined across latent classes. Categories that share
superscripts refer to comparisons that are not statistically different. Categories that do not share subscripts refer
to statistically significant difference p < .05. For all hypotheses tests, the alpha was set to 0.05. For the purposes
of display, the assigned most likely latent classes were cross-tabulated against the demographic criterion
variables.
98
After examining the latent class structure of this sample of students, we sought to
examine how the classes differed based on baseline demographic characteristics and risk
behaviors such as drug use and number of sexual partners. The associations based on
most likely class membership are presented in Table 11.
Based on this analysis, we find that the classes were both similar and different in
important ways. They did not differ significantly on ethnic composition (both comparing
all ethnicities and Latino vs. all else), parenthood, highest education level for parents,
having a job, and SADS. Nonetheless, the classes also differed demographically in
important ways. Classes 1 (M = 16.95, SD = 0.10) and 2 (M = 16.80, SD = 0.02) were
older then Class 3 (M = 16.39, SD = 0.09). Class 2 had a lower percentage of males
(56.40%) compared to Class 3 (73.20%).
The classes also differed in terms of certain risk taking behaviors. In terms of
thirty day substance use, Class 1 (M = 2.12, SD = 2.14) was found to have been drunk the
most times compared to Class 2 (M = 1.74, SD = 1.41) and Class 3 (M = 1.67, SD =
1.41). Class 3 used marijuana more times in the last thirty days (M = 3.61, SD = 3.84)
compared to Class 2 (M = 2.74, SD = 3.04). Not surprisingly, results regarding PEI (drug
consequences/problems) and 30 day number of sexual partners exhibited a similar pattern
such that Class 2 (M = 1.26, SD = 0.38) exhibited lower PEI scores than Class 3 (M =
1.35, SD = 0.54). In addition, Class 1 (M = 1.78, SD = 2.37) and Class 2 (M = 2.01, SD =
2.38) had less sexual partners than did individuals in Class 3 (M = 2.56, SD = 3.10).
99
Overall, a much simplified picture of the classes based on baseline composition
indicates that Class 3 was the youngest, had the most males, and also appeared to be the
most at risk – highest marijuana use, drug use/consequences (PEI) and largest number of
sexual partners. However, Class 1 had the highest rates of thirty day drunk on alcohol
compared to the other two classes.
Regression of Latent Classes on Outcomes
For calculation purposes, each subject was assigned to belong to their most likely
class. The estimated trajectory for the members in each of the classes is presented in
Figures 13-19. Table 11 provides descriptive information (means and frequencies) of all
Time 3 correlates by class.
100
Table 12. Class membership (N = 1,682) and endorsement frequencies of the Time 3 correlates using the best fitting LCGM solution
Class 1 Class 2 Class 3
Proportion
a
(N , %)
87, 5.27 1467, 88.86 97, 5.88
Mean SD Mean SD Mean SD
30-day cigarette use
b
1.90 2.57 2.49 3.10 2.39 2.84
30-day alcohol use
b
1.76 1.13 1.76 1.14 1.56 0.66
30-day drunk on alcohol
b
1.27 0.83 1.39 0.70 1.30 0.47
30-day marijuana use
b
1.92 2.34 1.88 2.08 1.87 2.34
Substance Abuse (SADS)
c
1.85 0.37 1.86 0.26 1.91 0.16
Drug Problems/Consequences (PEI)
d
1.04 0.24 1.08 0.22 1.09 0.2
30-day number of sexual parterns
f
2.02 2.33 1.90 1.48 1.82 1.11
% % %
Had a job 48.98 46.04 26.09
Was married 1.96 3.59 8.70
Was a parent 14.29 15.05 17.39
Condom used last intercourse
e
50.00 56.28 80.95
Note:
a
using most likely class,
b
response options were (1 = 0 times , 2 = 1-10 times , 3 = 11-20 times , 4 = 21-30 times , 5 = 31-40
times , 6 = 41-50 times , 7 = 51-60 times , 8 = 61-70 times , 9 = 71-80 times , 10 = 81-90 times , 11 = 91-100 times, 12 = over 100
times ),
c
(1 = yes , 2 = no ),
d
(1 = none , 2 = once or twice , 3 = sometimes , 4 = often ),
e
(1 = yes, 0 = no ),
f
(0 = 0 people , 1 = 1
person , 2 = 2 people , 3 = 3 people , 4 = 4 people , 5 = 5 people , 6 = 6 people , 7 = 7 people , 8 = 8 people , 9 = 9 people , 10 = 10
people , 11 = more than 10 people )
101
Mixed model analyses for categorical (PROC GLIMMIX) and continuous data
(PROC MIXED), using SAS version 9.2, were performed to analyze differences between
pairwise comparisons of emerging adulthood classes on Time 3 correlates. The results
are presented in Table 12. Latent class was not significantly associated with any of the
Time 3 outcomes except for having used a condom during the last intercourse, such that
Class 3 (80.95%) was significantly more likely to have used a condom compared to Class
1 (50.00%) (p < .05) and there was a significant trend between Class 3 and Class 2
(56.28%) (p = .06).
There was an additional statistically significant trend, such that Class 2 (M = 1.39,
SD = 0.70) was drunk the most times compared to Class 1 (M = 1.27, SD = 0.83; p =
.083) and compared to Class 3 (M = 1.30, SD = 0.47; p = .083).
102
Table 13. Differences for Time 3 outcomes by latent growth class.
Class 1 vs. Class 2 Class 1 vs. Class 3 Class 2 vs. Class 3 Overall
Estimate SE p Estimate SE p Estimate SE p F
30-day cigarette use
b
-0.122 0.121 0.315 -0.243 0.2064 0.2395 -0.121 0.175 0.4872 0.79
30-day alcohol use
b
-0.027 0.142 0.849 0.179 0.243 0.460 0.206 0.205 0.315 0.52
30-day drunk on alcohol
b
-0.259 0.149 0.083 -0.105 0.256 0.682 0.154 0.217 0.083 1.69
30-day marijuana use
b
0.046 0.138 0.741 -0.051 0.234 0.827 -0.097 0.198 0.624 0.17
Substance Abuse (SADS)
c
-0.047 0.147 0.748 -0.320 0.251 0.203 -0.273 0.212 0.199 0.90
Drug Problems/Consequences (PEI)
d
-0.141 0.137 0.306 -0.118 0.234 0.616 0.023 0.198 0.907 0.53
30-day number of sexual parterns
f
0.084 0.152 0.582 0.216 0.258 0.403 0.132 0.218 0.543 0.36
OR (95% CI) p OR (95% CI) p OR (95% CI) p
Had a job 1.04 (0.56 - 1.94) 0.900 0.43 (0.14 - 1.32) 0.140 0.41 (0.15 - 1.10) 0.075
Was married 3.14 (0.31 - 39.21) 0.312 8.81 (0.56 - 139.11) 0.122 2.97 (0.52 - 17.10) 0.222
Was a parent 0.74 (0.29 - 1.86) 0.518 0.82 (0.16 - 4.01) 0.806 1.10 (0.28 - 4.33) 0.891
Condom used last intercourse 1.55 (0.71 - 3.39) 0.270 5.11 (1.22 - 21.53) 0.026 3.27 (0.94 - 11.38) 0.063
103
Discussion
The aim of the present study was to examine the predictive utility of emerging
adulthood trajectories across three waves of data on behavioral outcomes. There are
numerous studies that examine substance use trajectories over time (Chassin et al., 2004;
Curran et al., 1997; Iwamoto et al., 2010; Stice, Barrera Jr, & Chassin, 1998). However,
many of these studies do not examine what longitudinal factors might impact drug use.
Studies that use cross sectional, two-wave designs fail to capture the full information that
is necessary to make inferences about changes over time and the predictive validity of
these studies is lacking. Non-linear variability, individually varying times of observation
and auto-regressive effects cannot be adequately ascertained without using true
longitudinal methodologies (Cheong et al., 2003; Rose et al., 2000).
The present study sought to use latent growth curve model as well as latent class
growth modeling to study the changes in emerging adulthood status over time.
Specifically, LGM is a promising technique as it allows for the modeling of growth and
individual differences over time (McArdle & Epstein, 1987; Meredith & Tisak, 1990).
This approach combines techniques from confirmatory factor analysis, repeated measures
multivariate analysis of variance, and structural equation modeling to analyze changes in
a given construct longitudinally. LGM is a valuable analysis technique as it is able to
effectively deal with common concerns such as missing data and model measurement
error. Thus, there are strong advantages for using LGM to model constructs over time
compared to using the more traditional fixed-effects techniques (Curran et al., 1997).
Because of the clear advantages of LGM and the paucity of studies that employ these
104
approaches to studying substance abuse over time in emerging adult populations, as well
as the absence of any studies examining the maturation process specifically over time, the
current study used LGM to model emerging adulthood status longitudinally.
LGM is a very valuable modeling technique and extends the possibilities of
longitudinal research exponentially. However, LGM is still limited in certain ways.
LGM assumes that all individuals in a sample come from the same population and that
the observed trajectory of growth is equally fitting for all the individuals in this
population. That is, for all individuals, the intercept is the same (identical starting point)
and the slope is the same (growth or decline) over time. To circumvent this assumption
one might use LCGM. LCGM postulates that the sample is not composed of individuals
from an identical population; instead the same sample is composed of more than one
population and that each population has a unique starting point (intercept) and growth
trajectory (slope). Group membership is a “latent” or unobserved construct (B. Muthen,
2001). The present study first identifies an overall sample trajectory using LGM, next it
uses the LCGM framework to identify varying developmental trends of latent classes
over time.
A growing number of studies have examined longitudinal substance use
trajectories during the transition from adolescence, through emerging adulthood and then
to adulthood (Anderson et al., 2010; Auerbach & Collins, 2006; Riggs et al., 2007; J
Schulenberg et al., 2005; J. S. Tucker et al., 2003; White et al., 2005).
The findings range in scope and provide an interesting and diverse picture of how
emerging adulthood looks in terms of substance use trajectories. Overall, substance use
105
appears to peak during emerging adulthood, but differs individually. This variation is
indicative of the heterogeneity of this period of the lifespan. One study found that non-
college attending individuals use more substances during the emerging adulthood period
than do their college-student peers (White et al., 2005). Those who engage in substance
use earlier tend to be at a higher risk for poorer outcomes later in life than those who do
not (Riggs et al., 2007; J. S. Tucker et al., 2005). This is just a small overview of
findings from longitudinal studies on substance use; there is still a lot to be done in this
arena. However, while a significant amount of the growth curve modeling research has
been completed on substance use during emerging adulthood, until now no study has
used LCGM to identify distinct trajectories of emerging adulthood as a construct over
time.
The first validation study of the IDEA (Reifman et al., 2007) examined how
emerging adulthood scores by sub-scale differed by age group. For the sub-scales
identity exploration, experimentation/possibilities, and negativity/instability (in both
samples) and self-focus scores were the highest in the 18 to 23 year old age group in
comparison to the older age groups. As expected, the reverse pattern was observed for
other-focus. Individuals in the emerging adulthood age range also scored higher on
identity exploration, other-focus, self-focus, and feeling “in-between” than did sixth
through twelfth graders. These results were evaluated in a cross-sectional sample and
thus required verification over a longer period of time. This longitudinal verification was
needed as without it is not possible to know whether these differences are due to cultural
and historical disparities across age cohorts rather than actual developmental processes.
106
Based on the first validation study of the IDEA (Reifman et al., 2007) and Study I
of the present dissertation it appears that the IDEA effectively captures emerging
adulthood as a developmental construct (indicated by higher scores). Lower scores on
the IDEA should represent either not yet having reached emerging adulthood or having
already resolved emerging adulthood (having moved on to adulthood). Thus, the present
study first used LGM to capture the overall trajectory shape as individuals pass through
emerging adulthood. As emerging adulthood is a period of great heterogeneity and is not
experienced by all individuals in the same manner LCGM was used to identify varying
groups of trajectories of emerging adulthood. These different populations or classes were
then regressed on correlates at Time 3 to determine whether how an individual passes
through emerging adulthood was predictive of various behavioral outcomes.
Factorial Invariance
The 8-item emerging adulthood scale was tested for factorial invariance across
time points. Measurement invariance is important for any type of longitudinal SEM
model regardless of number of time points. Factorial invariance over time can be
considered an empirical issue (Meredith & Horn, 2001). Within each occasion of
measurement is a latent variable or factor (f [1] with factor loadings λ
m
). Each variable
also has its own unique variation (Ψ
m
2
) indicated by double headed arrows in most path
diagrams. According to traditional factor analysis theory, unique factor scores can be
broken up into two distinct parts: (1) the part that is specific to the test and represents
valid measurement, and (2) random error. It is assumed that each factor contributed
107
variation at a given time point, but that each factor is also independent of other scores
within and across occasions of measurement (Meredith & Horn, 2001). Common factor
models are used to denote the testable hypothesis: a single unobserved latent variable can
account for covariation among the observed scores within each occasion of measurement.
The factor loadings for each must be the same value at all occasions, “factorial
invariance” such that λ
m
[1] = λ
m
[2] (or however many time points exist). This is a
statistical way of affirming that this factor score has the identical substantive meaning at
each time of measurement.
If factorial invariance is not achieved, it is possible to further relax constraints on
the model and test model fit to find a better fitting model; however, the results will likely
be difficult to interpret. If the number of factors changes between points of measurement,
then the factors would have to be interpreted separately and we could not assert that the
same common factors are being measured on each occasion. Repeated measures with the
same number of factors allow the researcher to claim invariant factor loadings and to
state the same constructs are repeated over time. Factorial invariance was indeed
achieved by eliminating three items from the scale. Thus, a 5-item emerging adulthood
measure was used for the subsequent analyses.
Latent Growth Curve Model
The latent growth curve model results indicate a sample that is gradually
increasing in emerging adulthood status from Time 1 to Time 3. As expected, there was
indeed a linear increase in emerging adulthood and the linear growth model fit the data
108
well both with and without covariates. However, the variance components for the
intercept and the slope indicate that there is significant individual variation. This is
indicative that not all individuals start at the same point on the IDEA and that not all
individuals grow at the same rate. Supportive of the contention that emerging adulthood
is an extremely heterogeneous period of life that is experienced differently by every
individual and allowing us to proceed with the latent growth model analysis.
Latent Class Growth Curve Model
As proposed, a LCGM analysis was conducted. The analysis extracted three
latent classes. The groups were very different sizes, Class 2 represented the largest group
(85.1%), and Class 1 (8.7%) and Class 3 (6.3%) were close in size. The shape of the
trajectory was different between classes for both intercept and slope; the groups were all
significantly different from each other. The steepest slope was found in Class 3, these
individuals increased linearly in emerging adulthood across the time points. Class 3
started the lowest on emerging adulthood, but had the steepest growth over time and
increased significantly from Time 1 to Time 3. This group was one of the smaller
groups, but followed more closely what is hypothesized for the transition to emerging
adulthood. Emerging adulthood is expected to increase from 18, peak, and then decline
until approximately age 25 when adulthood is reached (Arnett, 2004b; Reifman et al.,
2007). Class 2 was actually the largest group, with a very stable trajectory through time
on emerging adulthood (no growth); this class started the highest (intercept) on emerging
adulthood, but not did not increase or decrease significantly over time. Class 1 actually
109
started relatively high on emerging adulthood but declined over time. Had the sample
been normative youth, not continuation high school students, we might expect that the
groups would look a bit different. It’s possible that more normative youth would follow
the pattern of Class 3 by increasing rather steeply as they enter emerging adulthood.
Continuation high school youth are expected to have reached emerging adulthood earlier
than normative youth (Sussman, 2010) and this might explain why the largest group is
not characterized by steep growth.
The classes were compared on baseline demographic characteristics and risk
behaviors to get a picture of what these groups might look like. The classes did not differ
in terms of ethnic composition, parenthood, education level of parents, having a job, and
substance abuse. However, the classes did differ in other important ways.
Demographically, Class 1 and Class 2 were older than Class 3. Class 3 had a larger
percentage of males than did Class 2.
In terms of baseline risk behaviors important differences were also found between
classes. Specifically, for 30-day substance use, Class 3 used marijuana more times than
Class 2. Following in a similar fashion, Class 3 had the highest PEI scores compared to
Class 1 and Class 2. Class 1 and Class 2 had less sexual partners than Class 3.
Contrarily, Class 1 had been drunk more times than Class 2 and Class 3. Overall, we
might say that Class 3 appears to be the most at-risk based on baseline correlates.
110
Regression of Latent Class on Time 3 Correlates
Lastly, the preceding analyses examined each emerging adulthood trajectory
separately. The relationship of class membership on Time 3 outcomes was examined
controlling for baseline correlate score, age, gender, and ethnicity. It was believed that
the results would indicate, for example, whether emerging adults in Class 3 are more at
risk for problem behaviors such as cigarette or marijuana use at Time 3 than those in
Class 2. This might tell us that both intercept and growth rate was related to behavioral
outcomes later in life.
Unfortunately the results were not very informative. Only one statistically
significant difference was found between Class 1 and Class 3 on condom use during last
intercourse. Class 3 was more likely to have used a condom than Class 1. A statistically
significant trend emerged such that Class 2 was also less likely to have used a condom
than Class 3. While this finding could be important, it is suspected that we would find a
similar trend in number of sexual partners such that Class 3 was using a condom more
frequently because they had more sexual partners than members of Class 1 or Class 2
who might be in more stable monogamous relationships. As we did not find any
differences in terms of number of sexual partners, the information regarding condom use
is more difficult to interpret and might have simply emerged as a result of the multiple
comparisons.
A significant trend was found for times being drunk in the last thirty days, such
that Class 2 had been drunk more times than both Class 1 and Class 3. This finding is
111
difficult to interpret as no relationship is found between Class membership and times
having used alcohol, cigarettes nor marijuana.
Limitations
The limitations of this work include the use of a single continuation high school
sample and thus findings cannot be generalized to other populations. In addition, while
growth curve modeling makes use of FIML to account for missing data, there was still
substantial attrition over the study period. It should also be noted that LCGM yields
probabilities of class membership; thus the assignment of individuals to classes (while
model based) involves a degree of uncertainty that is not reflected in the Time 3
comparisons. Ideally the complete set of items used in Study I would have been used in
this study as this smaller measure has not been validated on its own. While factorial
invariance across time points was achieved using the 5-item scale, it would have been
best if the 8-item scale was factorially invariant over time. A foremost limitation is the
sample size of the individual latent classes. The latent trajectory groups were
imbalanced, and one class was considerably bigger than the other two groups. Thus,
there is a question of power in the ability to detect effects with the Time 3 correlates.
Lastly, there was not significant variation in the variance of the slope which might
indicate that the latent class growth curve model was not justified.
112
Conclusion
The present study sought to examine possible trajectories during emerging
adulthood. Three latent growth trajectories were extracted from the data. In terms of
intercepts and slopes, these trajectories differed in important ways. Class 1 started at
Time 1 relatively high (just a bit lower than Class 2) on emerging adulthood, but
decreased significantly over time. Class 2 was the largest and was relatively stable in
emerging adulthood over time. This group started the highest at Time 1 on emerging
adulthood and remained the highest across all time points. Class 3 began at the lowest
emerging adulthood score but rose significantly over time. In terms of Time 1 correlates,
Class 3 appeared to be the most at-risk. This group was characterized by the youngest
ages, the most males, and had used marijuana more times, scored highest on PEI and had
the highest number of 30-day sexual partners.
However, at Time 3 the differences between classes on the same correlates
examined at Time 1 appeared to be less apparent. There were significant differences
found in terms of condom use, such that Class 3 was most likely to use a condom during
the last intercourse. In addition, significant trends were found such that Class 2 was more
likely than the other classes to have been drunk more time times on alcohol. The results
should be interpreted in light of the limitation of the weak effect sizes and the multiple
uncorrected comparisons.
More importantly, this study provides verification that continuation high school
students are indeed entering emerging adulthood and do not simply bypass it by moving
directly to adulthood. The largest class, Class 2, appears to have entered emerging
113
adulthood at the start of the data collection, and stays in this period of life over the entire
date collection period. This should not be surprising as the approximate age of
participants at Time 3 is 18.7, which should be close to the highest point of experiencing
emerging adulthood according to the theory (Arnett, 2000b, 2004b; Reifman et al., 2007).
While Class 3 appears to be the most at risk at Time 1, the individuals in this class do not
differ on risk behaviors at Time 3. This might indicate that they successfully transitioned
into emerging adulthood as do more normative youth, and thus are able to negate their
early at-risk status.
The most surprising finding is the trajectory for Class 1. Individuals in this group
begin in emerging adulthood but decline consistently over time. We would think this
group might enter emerging adulthood early, and resolve it early as well by entering
adulthood. However, typical markers of having reached adulthood, such as having a job,
being married, or being a parent are not higher in this Class at Time 3 compared to the
other Classes.
Future Directions
Ideally, future studies will continue using this dataset as more time passes to get a
bigger picture of how these trajectory groups change over time. With an additional time
point we could investigate quadratic patterns for a more complete picture of the trajectory
through emerging adulthood. More research is needed that follows individuals over
longer periods of time and through various developmental stages.
114
CHAPTER 4: DISCUSSION
The overall objective of the present dissertation project was to elucidate the
etiology of drug use in Southern Californian continuation high school students as related
to emerging adulthood. In an attempt to accomplish this goal, this dissertation project
completed two studies that aim to explicate how a somewhat new construct, the IDEA,
advances what is already known about drug use etiology in continuation high school
students. Study I was used to determine the usefulness of the IDEA in a population of
Southern Californian continuation high school emerging adults. This study is the first
psychometric analysis of the IDEA in this population, thus providing valuable
information for future studies using this measure.
The IDEA was reduced to ten items in the last wave of data collection, thus, the
measure was used as a single factor measure and only the items used in all waves were
used in Study II. Study II attempts to examine the dynamic processes of emerging
adulthood with demographic characteristics as well as drug use behaviors over three
waves of data collection.
Results suggest that the IDEA can indeed be used in the current study population.
Continuation high school students, while younger than the standard conception of
emerging adulthood, indeed experience a transitional period in their lives indicative of
emerging adulthood. Correlational data at baseline indicates the IDEA is correlated with
drug use and other behaviors in this sample.
LCGM indicates that three developmentally distinct classes of emerging
adulthood are visible in the data. Contrary to expectation, these classes do not appear to
115
parallel typical substance use trajectory classes (Marti et al., 2010). These findings are
important to developmentally model and understand how individuals experience
emerging adulthood. This analysis allows us to gain a better understanding of the true
heterogeneity that characterizes this complex developmental period.
This analysis not only provides valuable information for drug use prevention, but
it also allows the theory of emerging adulthood (Arnett, 2000b) to be advanced by
determining possible trajectories of how individuals pass through emerging adulthood.
Not all individuals experience emerging adulthood in the same way. Latent classes of
emerging adulthood emerged and thus subsequent inferences can be made about how
individuals experience this discrete period of life. It appears that individuals enter
emerging adulthood at varying ages, and once they have entered it they “grow” at
different paces.
Specifically, two classes (2 and 3) were found to enter emerging adulthood earlier
than one might expect in a non-continuation high school student population. Class 2, the
largest class, remains in emerging adulthood throughout the study. This indicates a
population who has not yet resolved emerging adulthood; supported by their young age
(M = 19.7) even at the end of the data collection period. Another latent growth class,
Class 3, is just entering emerging adulthood during the period studied. This group might
be what we would expect in a more normative group of individuals who might enter
emerging adulthood around 18 years of age, and gradually increase, peak, and decline
until they resolve at approximately 25 years of age. Lastly, Class 1, possibly the most
interesting class because of the unexpected trajectory it follows, entered emerging
116
adulthood quite early, but declines significantly over time. However, the Time 3
outcomes (parenthood, marriage, etc.) do not lend support as to whether this group has
actually resolved emerging adulthood and entered young adulthood.
Limitations
The dissertation studies were not without limitations. First for Study I replication
is necessary to fully validate any measure. Second, cross-sectional data were used, as
such; replication with longitudinal data using the full IDEA would be helpful to enhance
our understanding of changes on the IDEA over time, as well as the order of precedence
between its factors and other variables. In addition, various dissimilarities in sample
characteristics (age, gender, ethnicity) might explain differences between the original
(Reifman et al., 2007) and current psychometric evaluation of the IDEA. Fourth, the
subjects were continuation high school students, so study results cannot be generalized
outside of this population. Moreover, ideally the full IDEA scale with all six subscales
should be validated. Finally, the subscales are not completely distinct; however, this
might simply indicate that the subscales are under the umbrella of the overarching
concept of emerging adulthood.
Study II must also be evaluated with certain limitations in mind. This study
began with a limited number of items from the IDEA measure. It might have been nice
to have the complete set of items used in Study I as this smaller number has not been
validated on its own. While factorial invariance across time points was achieved using
the 5-item scale, it would have been ideal if the 8-item scale held up across time. A
117
major concern is the sample size of the individual latent classes. The latent trajectory
groups were unequal, and one class was substantially larger than the other two groups.
Thus, there is a question of power in the ability to detect effects with the Time 3
correlates. As with Study I, the sample used was a continuation high school population
and thus results cannot be generalized outside of this population. In addition, LCGM
yields probabilities of class membership; thus the assignment of individuals to classes
(although model based) comprises a degree of uncertainty that is not reflected in the Time
3 comparisons. Lastly, while growth curve modeling allows for rigorous accounting of
missing data and the full data is used through FIML, it cannot be ignored that there was
indeed substantial attrition of the sample over time.
Implications for Future Research
Ideally the findings from this dissertation project will provide important
implications for alcohol and drug use etiologies and can be used to improve drug
prevention programs directed at at-risk populations such as continuation high school
students in Southern California. As it appears that this at-risk sample is entering
emerging adulthood earlier than other normative students, it might be important to target
these students at an earlier age for alcohol and drug use prevention. Future research
might determine whether emerging adulthood is a malleable construct. If it is, then drug
prevention efforts might target the specific dimensions of this construct to help deter
individuals from experiencing hazardous transitions into adulthood. It would be
118
interesting to compare normative youth to continuation high school youth using a similar
analysis.
It is important to continue to conduct long-term longitudinal studies that follow
youth from adolescence, to emerging adulthood, to young adulthood. It is only by
following the same sample through multiple developmental periods that we can gain a
clearer understanding of how developmental trajectories affect future outcomes such as
health, education, substances use and other psychosocial and behavioral outcomes later in
life. Research has only found limited success (J. S. Tucker et al., 2005) in identifying
determinants and patterns of substance use over time. Identifying and better
understanding the factors that impact use must continue to be researched. These factors
might put forward novel and useful approaches for decreasing initiation and acceleration
of substance use between adolescence and emerging adulthood.
119
BIBLIOGRAPHY
Anderson, K. G., Ramo, D. E., Cummins, K. M., & Brown, S. A. (2010). Alcohol and
drug involvement after adolescent treatment and functioning during emerging
adulthood. Drug and alcohol dependence, 107(2-3), 171-181.
Arias, D. F., & Hernandez, A. M. (2007). Emerging Adulthood in Mexican and Spanish
Youth: Theories and Realities. Journal of Adolescent Research, 22(5), 28.
Arnett, J. J. (1997). Young people's conceptions of the transition to adulthood. Youth &
Society, 29(1), 3.
Arnett, J. J. (1998). Learning to stand alone: The contemporary American transition to
adulthood in cultural and historical context. Human Development, 41(5-6), 295-
315.
Arnett, J. J. (2000a). Adolescence and emerging adulthood: Prentice Hall.
Arnett, J. J. (2000b). Emerging adulthood - A theory of development from the late teens
through the twenties. American Psychologist, 55(5), 469-480.
Arnett, J. J. (2004a). Adolescence and emerging adulthood: A cultural approach (2nd
ed.). Upper Saddle River, New Jersey: Pearson.
Arnett, J. J. (2004b). Emerging adulthood: The winding road from the late teens through
the twenties: Oxford University Press, USA.
Arnett, J. J. (2005). The developmental context of substance use in emerging adulthood.
Journal of drug issues, 35(2), 235.
Arnett, J. J. (2006). Emerging adulthood in Europe: A response to Bynner. Journal of
youth studies, 9(1), 111-123.
Arnett, J. J. (2007a). Afterword: Aging out of care-Toward realizing the possibilities of
emerging adulthood. New Directions for Youth Development, 2007(113), 151-
161.
Arnett, J. J. (2007b). Emerging Adulthood [Special Issue]. Journal of Adolescent
Research, 22(5).
Arnett, J. J., & Tanner, J. L. (2006). Emerging adults in America: coming of age in the
21st century: Amer Psychological Assn.
120
Atak, H., & Çok, F. (2008). The Turkish Version of Inventory of the Dimensions of
Emerging Adulthood (The IDEA). International Journal of Human and Social
Sciences, 2(3).
Auerbach, M., & Collins, L. (2006). A Multidimensional Developmental Model of
Alcohol Use During Emerging Adulthood. Journal of Studies on Alcohol and
Drugs, 917-925.
Bachman, J. G., Wadsworth, K. N., Johnston, L. D., & Schulenberg, J. E. (1997).
Smoking, Drinking, and Drug Use in Young Adulthood: The Impacts of New
Freedoms and New Responsibilities. Research Monographs in Adolescence
(RMA): Lawrence Erlbuam Associates, Inc.
Bennett, M. E., McCrady, B. S., Johnson, V., & Pandina, R. J. (1999). Problem Drinking
from Young Adulthood to Adulthood: Patterns, Predictors and Outcomes (*).
Journal of Studies on Alcohol, 60(5).
Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological
bulletin, 107(2), 238.
Blackson, T. C., Butler, T., Belsky, J., Ammerman, R. T., Shaw, D. S., & Tarter, R. E.
(1999). Individual traits and family contexts predict sons' externalizing behavior
and preliminary relative risk ratios for conduct disorder and substance use
disorder outcomes. Drug and alcohol dependence, 56(2), 115-131.
Bollen, K. (1989). Structural Equations With Latent Variables. New York: Wiley.
Brook, J., Adams, R., Balka, E., & Johnson, E. (2002). Early adolescent marijuana use:
Risks for the transition to young adulthood. Psychological Medicine, 32(01), 79-
91.
Brook, J., Whiteman, M., Cohen, P., Shapiro, J., & Balka, E. (1995). Longitudinally
predicting late adolescent and young adult drug use: childhood and adolescent
precursors. Journal of the American Academy of Child & Adolescent Psychiatry,
34(9), 1230-1238.
Brown, R. (1989). Using covariance modeling for estimating reliability on scales with
ordered polytomous variables. Educational and Psychological Measurement,
49(2), 385.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. Testing
structural equation models, 154, 136–162.
121
Buhl, H. M., & Lanz, M. (2007). Emerging Adulthood in Europe. Journal of Adolescent
Research, 22(5), 439.
Bumpass, L., & Lu, H. (2000). Trends in cohabitation and implications for children s
family contexts in the United States. Population Studies, 54(1), 29-41.
Burns, A., & Dunlop, R. (1998). Parental divorce, parent–child relations, and early adult
relationships: A longitudinal Australian study. Personal relationships, 5(4), 393-
407.
Byrne, B. M., Shavelson, R. J., & Muthén, B. (1989). Testing for the equivalence of
factor covariance and mean structures: The issue of partial measurement
invariance. Psychological Bulletin, 105(3), 456.
Campbell, K., Zobeck, T., & Bertolucci, D. (1996). Trends in alcohol-related fatal traffic
crashes, United States, 1977-1993: NIAAA Surveillance Report.
Chassin, L., Flora, D. B., & King, K. M. (2004). Trajectories of alcohol and drug use and
dependence from adolescence to adulthood: The effects of familial alcoholism
and personality. Journal of Abnormal Psychology, 113, 483-498.
Chassin, L., Pitts, S. C., & Prost, J. (2002). Binge drinking trajectories from adolescence
to emerging adulthood in a high-risk sample: Predictors and substance abuse
outcomes. Journal of Consulting and Clinical Psychology, 70(1), 67-78.
Chassin, L., Presson, C. C., Pitts, S. C., & Sherman, S. J. (2000). The Natural History of
Cigarette Smoking From Adolescence to Adulthood in a Midwestern Community
Sample: Multiple Trajectories and Their Psychosocial Correlates* 1. Health
Psychology, 19(3), 223-231.
Chen, K., & Kandel, D. B. (1995). The natural history of drug use from adolescence to
the mid-thirties in a general population sample. American journal of public
health, 85(1), 41.
Cheong, J. W., MacKinnon, D. P., & Khoo, S. T. (2003). Investigation of mediational
processes using parallel process latent growth curve modeling. Structural
equation modeling: a multidisciplinary journal, 10(2), 238-262.
Cheung, G. W., & Rensvold, R. B. (2000). Assessing extreme and acquiescence response
sets in cross-cultural research using structural equations modeling. Journal of
Cross-Cultural Psychology, 31(2), 187.
Cloninger, C. R. (1987). Neurogenetic adaptive mechanisms in alcoholism. Science,
236(4800), 410.
122
Cohen, S. (1988). Perceived stress in a probability sample of the United States.
Cohen, S., Kamarck, T., & Mermelstein, R. (1983). A global measure of perceived stress.
Journal of health and social behavior, 24(4), 385-396.
Cole, D. A., & Maxwell, S. E. (2003). Testing mediational models with longitudinal data:
Questions and tips in the use of structural equation modeling. Journal of
Abnormal Psychology, 112(4), 558-577.
Collins, L. M., Schafer, J. L., & Kam, C. M. (2001). A comparison of inclusive and
restrictive strategies in modern missing data procedures. Psychological Methods,
6(4), 330-351.
Commendador, K. (2007). The relationship between female adolescent self esteem,
decision making, and contraceptive behavior. Journal of the American Academy
of Nurse Practitioners, 19(11), 614-623.
Curran, P. J. (2000). A latent curve framework for the study of developmental trajectories
in adolescent substance use. Multivariate applications in substance use research:
New methods for new questions, 1-42.
Curran, P. J., Harford, T. C., & Muthen, B. O. (1996). The Relation between Heavy
Alcohol Use and Bar Patronage: A Latent Growth Model. Journal of Studies on
Alcohol, 57(4).
Curran, P. J., Stice, E., & Chassin, L. (1997). The relation between adolescent alcohol use
and peer alcohol use: A longitudinal random coefficients model. Journal of
Consulting and Clinical Psychology, 65(1), 130.
Côté, J. (2000). Arrested adulthood: The changing nature of maturity and identity: NYU
Press.
Dawson, D. A. (2000). The link between family history and early onset alcoholism:
earlier initiation of drinking or more rapid development of dependence? Journal
of Studies on Alcohol and Drugs, 61(5), 637.
de Moor, C., Johnston, D., Werden, D., Elder, J., Senn, K., & Whitehorse, L. (1994).
Patterns and correlates of smoking and smokeless tobacco use among
continuation high school students. Addictive behaviors, 19(2), 175-184.
Deng, X., Doll, W. J., Hendrickson, A. R., & Scazzero, J. A. (2005). A multi-group
analysis of structural invariance: an illustration using the technology acceptance
model. Information & management, 42(5), 745-759.
123
Dennis, M., Titus, J., White, M., Unsicker, J., & Hodgkins, D. (2002). Global Appraisal
of Individual Needs (GAIN): Administration guide for the GAIN and related
measures. Bloomington, IL: Chestnut Health Systems.
DeWit, D. J. (1998). Frequent childhood geographic relocation:: Its impact on drug use
initiation and the development of alcohol and other drug-related problems among
adolescents and young adults. Addictive Behaviors, 23(5), 623-634.
Drasgow, F. (1984). Scrutinizing psychological tests: Measurement equivalence and
equivalent relations with external variables are the central issues.
Drasgow, F., & Kanfer, R. (1985). Equivalence of psychological measurement in
heterogeneous populations. Journal of Applied Psychology, 70(4), 662.
Duris, J. (2009). Erleben des Erwachsenwerdens und Bewertung von alltäglichen
Risikosituationen.
Ellickson, P. L., Tucker, J. S., & Klein, D. J. (2001). Sex differences in predictors of
adolescent smoking cessation. Health Psychology, 20(3), 186.
Erikson, E. (1950). Childhood and Society. New York: Norton.
Erikson, E. (1959). Identity and the life cycle: Psychological issues. New York: Int. Univ.
Press (trad. it.: I cicli della vita. Continuità e mutamenti. Roma: Armando, 1991).
Facio, A., Resett, S., Micocci, F., & Mistrorigo, C. (2007). Emerging Adulthood in
Argentina: An age of diversity and possibilities. Child Development Perspectives,
1(2), 115-118.
Fleming, C. B., Mason, W. A., Mazza, J. J., Abbott, R. D., & Catalano, R. F. (2008).
Latent growth modeling of the relationship between depressive symptoms and
substance use during adolescence. Psychology of Addictive Behaviors, 22(2), 186-
197.
Franklin, C., & Streeter, C. L. (1995). Assessment of middle class youth at-risk to
dropout: School, psychological and family correlates. Children and Youth
Services Review, 17(3), 433-448.
Fry, R. (2009). College Enrollment Hits All-Time High, Fueled by Community College
Surge - Pew Research Center, from http://pewresearch.org/pubs/1391/college-
enrollment-all-time-high-community-college-surge
Furstenberg, F. F., Brooks-Gunn, J., & Morgan, S. P. (1989). Adolescent mothers in later
life: Cambridge Univ Pr.
124
Goldscheider, F., & Goldscheider, C. (1999). The changing transition to adulthood: Sage
Publications.
Gollubits, S. (2010). Wird das Erwachsenwerden in Abhängigkeit von der Ausbildung
unterschiedlich erlebt?
Gorski, A. (2009). From freedom fighters to governors.
Graham, J. W., Flay, B. R., Johnson, C. A., Hansen, W. B., Grossman, L., & Sobel, J. L.
(1984). Reliability of self-report measures of drug use in prevention research:
Evaluation of the Project SMART questionnaire via the test-retest reliability
matrix. Journal of drug education, 14(2), 175-193.
Greenbaum, P. E., & Dedrick, R. F. (2007). Changes in Use of Alcohol, Marijuana, and
Services by Adolescents With Serious Emotional Disturbance. Journal of
Emotional and Behavioral Disorders, 15(1), 21.
Griffin, C. (1993). Representations of youth: The study of youth and adolescence in
Britain and America: Polity Press.
Grunbaum, J., Kann, L., Kinchen, S., Ross, J., Gowda, V., Collins, J., & Kolbe, L.
(2000). Youth risk behavior surveillance national alternative high school youth
risk behavior survey, United States, 1998. Journal of School Health, 70(1), 5-17.
Hager, B. (2009). Wie wird Erwachsenwerden erlebt?
Hawkins, J. D., Catalano, R. F., & Miller, J. Y. (1992). Risk and protective factors for
alcohol and other drug problems in adolescence and early adulthood: implications
for substance abuse prevention. Psychological bulletin, 112(1), 64.
Hawkins, J. D., Graham, J. W., Maguin, E., Abbott, R., Hill, K. G., & Catalano, R. F.
(1997). Exploring the effects of age of alcohol use initiation and psychosocial risk
factors on subsequent alcohol misuse. Journal of studies on alcohol, 58(3), 280.
Hayes, C. D. (1987). Risking the future: Adolescent sexuality, pregnancy, and
childbearing: National Academies.
Heckhausen, J. (1999). Developmental regulation in adulthood: Age-normative and
sociostructural constraints as adaptive challenges: Cambridge Univ Pr.
Herrera, D. E. (2009). Perceived racial/ethnic discrimination, hope, and social
connectedness: examining the predictors of future orientation among emerging
adults.
125
Holahan, C., Pasch, K., & Steinhardt, M. The Dissertation Committee for Denise E.
Herrera certifies that this is the approved version of the following dissertation.
Holahan, C., Pasch, K., & Steinhardt, M. (2009). The Dissertation Committee for Denise
E. Herrera certifies that this is the approved version of the following dissertation.
Hornblower, M. (1997). Great xpectations. Time, 149(23), 58-69.
Horowitz, A. D., & Bromnick, R. D. (2007). Contestable Adulthood. Youth & Society,
39(2), 209.
Hurley, A. E., Scandura, T. A., Schriesheim, C. A., Brannick, M. T., Seers, A.,
Vandenberg, R. J., & Williams, L. J. (1997). Exploratory and confirmatory factor
analysis: Guidelines, issues, and alternatives. Journal of Organizational Behavior,
18(6), 667-683.
Iwamoto, D. K., Corbin, W., & Fromme, K. (2010). Trajectory classes of heavy episodic
drinking among Asian American college students. Addiction, 105(11), 1912-1920.
Jackson, K. M., Sher, K. J., Rose, R. J., & Kaprio, J. (2009). Trajectories of Tobacco Use
from Adolescence to Adulthood: Are the Most Informative Phenotypes Tobacco
Specific? National Cancer Institute. Phenotypes and Endophenotypes:
Foundations for Genetic Studies of Nicotine Use and Dependence. Tobacco
Control Monograph No, 20.
Jackson, K. M., Sher, K. J., & Wood, P. K. (2000). Trajectories of concurrent substance
use disorders: A developmental, typological approach to comorbidity.
Alcoholism: Clinical and Experimental Research, 24(6), 902-913.
Johnston, L., Bachman, J., & Schulenberg, J. (2005). Monitoring the Future: National
Survey Results on Drug Use, 1975-2004. Volume I: Secondary School Students,
2004. National Institutes of Health, 700.
Johnston, L., Bachman, J., & Schulenberg, J. (2009). Monitoring the Future: National
Survey Results on Drug Use, 1975-2008. Volume I, Secondary School Students.
NIH Publication No. 09-7402. National Institute on Drug Abuse (NIDA), 758.
Justice, N. I. o. (1996). Drug Use Forecasting 1995: Annual Report on Adult and Juvenile
Arrestees: National Institute of Justice.
Kandel, D., Simcha-Fagan, O., & Davies, M. (1986). Risk factors for delinquency and
illicit drug use from adolescence to young adulthood. Journal of Drug Issues.
126
Kandel, D. B., & Chen, K. (2000). Types of marijuana users by longitudinal course.
Journal of Studies on Alcohol, 61(3), 367.
Kandel, D. B., Kessler, R. C., & Margulies, R. Z. (1978). Antecedents of adolescent
initiation into stages of drug use: A developmental analysis. Journal of Youth and
Adolescence, 7(1), 13-40.
Kreyszig, S. (2007). Imaginary audience and voice in undergraduate emerging adults:
Library and Archives Canada= Bibliothèque et Archives Canada.
Krohn, M., Lizotte, A., & Perez, C. (1997). The interrelationship between substance use
and precocious transitions to adult statuses. Journal of Health and Social
Behavior, 38(1), 87-103.
Lanza, S. T., Patrick, M. E., & Maggs, J. L. (2010). Latent Transition Analysis: Benefits
of a Latent Variable Approach to Modeling Transitions in Substance Use. Journal
of drug issues, 40(1), 93.
Larson, R. (1990). The solitary side of life: An examination of the time people spend
alone from childhood to old age. Developmental Review, 10(2), 155-183.
Lawrence, F. R., & Hancock, G. R. (1998). Assessing Change over Time Using Latent
Growth Modeling. Measurement and Evaluation in Counseling and Development,
30(4), 211-224.
Lisha, N. E., Sun, P., Rohrbach, L., Spruijt-Metz, D., Unger, B., & Sussman, S. (in
press). An evaluation of immediate outcomes and fidelity of a drug abuse
prevention program in continuation high schools: Project Towards No Drug
Abuse (TND). Journal of Drug Education.
Little, R. J. A., & Rubin, D. B. (1987). Statistical analysis with missing data.
Lo, Y., Mendell, N. R., & Rubin, D. B. (2001). Testing the number of components in a
normal mixture. Biometrika, 88(3), 767-778.
Long, B. C. (1998). Coping with workplace stress: A multiple-group comparison of
female managers and clerical workers. Journal of Counseling Psychology, 45(1),
65.
Luyckx, K., Goossens, L., & Soenens, B. (2006). A developmental contextual perspective
on identity construction in emerging adulthood: Change dynamics in commitment
formation and commitment evaluation. Developmental Psychology, 42(2), 366.
127
Macek, P., Bej ek, J., & Vaní ková, J. (2007). Contemporary Czech Emerging Adults.
Journal of adolescent research, 22(5), 444.
Marti, C. N., Stice, E., & Springer, D. W. (2010). Substance use and abuse trajectories
across adolescence: A latent trajectory analysis of a community-recruited sample
of girls. Journal of Adolescence, 33(3), 449-461.
Marzana, D., Pérez-Acosta, A. M., Marta, E., & González, M. I. (2010). La transición a
la edad adulta en Colombia: una lectura relacional. Avances en Psicología
Latinoamericana, 28(1), 99-112.
Mayer, K. (2004). Whose lives? How history, societies, and institutions define and shape
life courses. Research in Human Development, 1(3), 161-187.
Mayseless, O., & Scharf, M. (2003). What does it mean to be an adult? The Israeli
experience. New directions for child and adolescent development, 2003(100), 5-
20.
McArdle, J. J., & Epstein, D. (1987). Latent growth curves within developmental
structural equation models. Child Development, 58(1), 110-133.
McArdle, J. J., & Hamagami, F. (1992). Modeling incomplete longitudinal and cross-
sectional data using latent growth structural models. Experimental Aging
Research.
McLachlan, G. J., & Peel, D. (2000). Finite mixture models (Vol. 299): Wiley-
Interscience.
McRobbie, A. (1991). Feminism and youth culture: from" Jackie" to" Just seventeen":
Macmillan.
Meredith, W., & Horn, J. (2001). The role of factorial invariance in modeling growth and
change.
Meredith, W., & Tisak, J. (1990). Latent curve analysis. Psychometrika, 55(1), 107-122.
Miniño, A. M. (2010). Mortality among teenagers aged 12-19 years: United States, 1999-
2006. NCHS data brief, no 37. In M. N. C. f. H. S. Hyattsville (Ed.).
Moscicki, E. (1995). Epidemiology of suicide. International Psychogeriatrics, 7(02),
137-148.
Murray, D. M., & Hannan, P. J. (1990). Planning for the appropriate analysis in school-
based drug-use prevention studies. Journal of Consulting and Clinical
Psychology, 58(4), 458-468.
128
Murray, D. M., Varnell, S. P., & Blitstein, J. L. (2004). Design and analysis of group-
randomized trials: a review of recent methodological developments. American
Journal of Public Health, 94(3), 423.
Musante, D. S. (2010). Family Predictors of Negative Instability in Adopted Emerging
Adults. University of Massachusetts Amherst.
Muthen, B. (2001). Latent variable mixture modeling. New developments and techniques
in structural equation modeling, 1–33.
Muthen, B., Brown, C. H., Masyn, K., Jo, B., Khoo, S. T., Yang, C. C., . . . Liao, J.
(2002). General growth mixture modeling for randomized preventive
interventions. Biostatistics, 3(4), 459.
Muthen, L., & Muthen, B. (2007). MPlus User's guide Fifth edition. Muthén & Muthén.
Muthén, B. (2001). Latent variable mixture modeling. New developments and techniques
in structural equation modeling, 1–33.
Muthén, B. (2002). Beyond SEM: General latent variable modeling. Behaviormetrika,
29(1; ISSU 51), 81-118.
Muthén, B. (2004). Latent variable analysis: Growth mixture modeling and related
techniques for longitudinal data. Handbook of quantitative methodology for the
social sciences, 345–368.
Muthén, B., & Shedden, K. (1999). Finite mixture modeling with mixture outcomes using
the EM algorithm. Biometrics, 55(2), 463-469.
Muthén, L., & Muthén, B. (2007). MPlus User's guide Fifth edition. Muthén & Muthén.
Nagin, D. S. (1999). Analyzing developmental trajectories: A semiparametric, group-
based approach. Psychological methods, 4, 139-157.
Newcomb, M., & Bentler, P. (1988). Consequences of adolescent drug use: Impact on the
lives of young adults. No.: ISBN 0-8039-2847-5, 285.
NIAAA. (1997). Alcohol and Health. Ninth Special Report to the U.S. Congress from the
Secretary of Health and Human Services.: National Inst. on Alcohol Abuse and
Alcoholism (DHHS), Rockville, MD.
NPR.). Marrying age in the United States Retrieved February 1st, 2011, from
http://www.npr.org/news/graphics/2009/jun/marriage/
129
Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the number of
classes in latent class analysis and growth mixture modeling: A Monte Carlo
simulation study. Structural Equation Modeling: A Multidisciplinary Journal,
14(4), 535-569.
Phinney, J. (1999). Ethnic families in Southern California: Change and stability in
multicultural settings.
Raftery, A. E. (1995). Bayesian model selection in social research. Sociological
methodology, 25, 111-164.
Reifman, A., Arnett, J. J., & Colwell, M. J. (2007). Emerging adulthood: Theory,
assessment, and application. Journal of Youth Development, 2(1).
Riggs, N. R., Chou, C. P., Li, C., & Pentz, M. A. (2007). Adolescent to emerging
adulthood smoking trajectories: When do smoking trajectories diverge, and do
they predict early adulthood nicotine dependence? Nicotine & tobacco research,
9(11), 1147.
Rindfuss, R., Swicegood, C., & Rosenfeld, R. (1987). Disorder in the life course: How
common and does it matter? American Sociological Review, 785-801.
Rogosa, D. (1987). Casual models do not support scientific conclusions: A comment in
support of Freedman. Journal of Educational Statistics, 12(2), 185-195.
Rogosa, D. (1988). Myths about longitudinal research.
Rohde, P., Lewinsohn, P. M., Brown, R. A., Gau, J. M., & Kahler, C. W. (2003).
Psychiatric disorders, familial factors and cigarette smoking: I. Associations with
smoking initiation. Nicotine & tobacco research, 5(1), 85-98.
Rohrbach, L., Sussman, S., Dent, C., & Sun, P. (2005). Tobacco, alcohol, and other drug
use among high-risk young people: A five-year longitudinal study from
adolescence to emerging adulthood. Journal of Drug Issues, 35(2), 333-355.
Rose, J. S., Chassin, L., Presson, C. C., & Sherman, S. J. (2000). Multivariate
applications in substance use research: New methods for new questions:
Lawrence Erlbaum.
Rosenthal, D. A., Gurney, R. M., & Moore, S. M. (1981). From trust on intimacy: A new
inventory for examining erikson's stages of psychosocial development. Journal of
Youth and Adolescence, 10(6), 525-537.
Rubin, D. B. (1976). Inference and missing data. Biometrika, 63(3), 581.
130
Sampson, R., & Laub, J. (1995). Crime in the making: Pathways and turning points
through life: Harvard Univ Pr.
SAS. (2008). SAS/STAT® 9.2 User's Guide.
Scheffer, J. (2002). Dealing with missing data. Research letters in the information and
mathematical sciences, 3(1), 153-160.
Schoonbroodt, A., & Jones, L. (2007). Baby busts and baby booms: the fertility response
to shocks in dynastic models.
Schulenberg, J., Merline, A. C., Johnston, L. D., O'Malley, P. M., Bachman, J. G., &
Laetz, V. B. (2005). Trajectories of marijuana use during the transition to
adulthood: The big picture based on national panel data. Journal of drug issues,
35(2), 255.
Schulenberg, J., O'Malley, P. M., Bachman, J. G., Wadsworth, K. N., & Johnston, L. D.
(1996). Getting drunk and growing up: Trajectories of frequent binge drinking
during the transition to young adulthood. Journal of Studies on Alcohol and
Drugs, 57(3), 289.
Schulenberg, J., Wadsworth, K. N., O'Malley, P. M., Bachman, J. G., & Johnston, L. D.
(1996). Adolescent Risk Factors for Binge Drinking During the Transition to
Young Adulthood: Variable-and Pattern-Centered Approaches to Change* 1.
Developmental Psychology, 32(4), 659-674.
Schwartz, S. J., Côté, J. E., & Arnett, J. J. (2005). Identity and agency in emerging
adulthood. Youth & Society, 37(2), 201.
Schwarz, G. (1978). Estimating the dimension of a model. The annals of statistics, 461-
464.
Schönhart, S. M. (2009). Bindung und Erwachsen werden.
Settersten Jr, R. A., & Ray, B. (2010). What's going on with young people today? The
long and twisting path to adulthood. The future of children, 19-41.
Sher, K. J., Grekin, E. R., & Williams, N. A. (2005). The development of alcohol use
disorders. Clinical Psychology, 1(1), 493.
Shulman, S., & Ben-Artzi, E. (2003). Age-related differences in the transition from
adolescence to adulthood and links with family relationships. Journal of Adult
Development, 10(4), 217-226.
131
Siddiqui, O., Hedeker, D., Flay, B. R., & Hu, F. B. (1996). Intraclass correlation
estimates in a school-based smoking prevention study. American Journal of
Epidemiology, 144(4), 425.
Singer, J. D. (1998). Using SAS PROC MIXED to fit multilevel models, hierarchical
models, and individual growth models. Journal of Educational and Behavioral
Statistics, 23(4), 323.
Sirsch, U., Dreher, E., Mayr, E., & Willinger, U. (2009). What Does It Take to Be an
Adult in Austria? Journal of Adolescent Research, 24(3), 275.
Stam, H., Hartman, E. E., Deurloo, J. A., Groothoff, J., & Grootenhuis, M. A. (2006).
Young adult patients with a history of pediatric disease: impact on course of life
and transition into adulthood. Journal of adolescent health, 39(1), 4-13.
Stice, E., Barrera Jr, M., & Chassin, L. (1998). Prospective differential prediction of
adolescent alcohol use and problem use: Examining the mechanisms of effect.
Journal of Abnormal Psychology, 107(4), 616.
Stoolmiller, M., Duncan, T., Bank, L., & Patterson, G. R. (1993). Some problems and
solutions in the study of change: Significant patterns in client resistance. Journal
of Consulting and Clinical Psychology, 61(6), 920.
Supper, B. (2009). Die Bedeutung der Persönlichkeit im Übergang zum
Erwachsenenalter.
Sussman, S. (2010). Emerging Adulthood and Substance Abuse. In L. V. Berhardt (Ed.),
Advances in Medicine and Biology (Vol. 6, pp. 221-231): Nova Science
Publishers.
Sussman, S., Dent, C., & Leu, L. (2000). The one-year prospective prediction of
substance abuse and dependence among high-risk adolescents. Journal of
Substance Abuse, 12(4), 373-386.
Sussman, S., Dent, C., & Stacy, A. (2002). Project Towards No Drug Abuse: A review of
the findings and future directions. American Journal of Health Behaviour, 26(5),
354-365.
Sussman, S., Rohrbach, L., Skara, S., & Dent, C. (2004). Prospective Prediction of
Alternative High School Graduation Status at Emerging Adulthood. Journal of
Applied Social Psychology, 34(12), 2452-2468.
132
Sussman, S., Stacy, A., Dent, C., Simon, T., Galaif, E., Moss, M., . . . Johnson, C. (1995).
Continuation high schools: Youth at risk for drug abuse. Journal of drug
education, 25(3), 191-209.
Sussman, S., Sun, P., Rohrbach, L., & Spruijt-Metz, D. (under review). One-year
outcomes of a drug abuse prevention program for older teens: Evaluating a
motivational interviewing booster component. Health Psychology.
Swisher, L., Beckstead, J., & Bebeau, M. (2004). Factor analysis as a tool for survey
analysis using a professional role orientation inventory as an example. Physical
Therapy, 84(9), 784.
Tan, X., Dierker, L., Rose, J., & Li, R. (2011). How Spacing of Data Collection May
Impact Estimates of Substance Use Trajectories. Substance Use & Misuse.
Thornberry, T. P. (1987). Toward an interactional theory of delinquency. Criminology,
25(4), 863-892.
Tucker, J. S., Ellickson, P. L., Orlando, M., Martino, S. C., & Klein, D. J. (2005).
Substance Use Trajectories From Early Adolescence to Emerging Adulthood: A
Comparison of Smoking, Binge Drinking, and Marijuana Use. Journal of Drug
Issues.
Tucker, J. S., Orlando, M., & Ellickson, P. L. (2003). Patterns and Correlates of Binge
Drinking Trajectories From Early Adolescence to Young Adulthood* 1. Health
Psychology, 22(1), 79-87.
Tucker, L. R., & Lewis, C. (1973). A reliability coefficient for maximum likelihood
factor analysis. Psychometrika, 38(1), 1-10.
Tuinstra, J., Van Sonderen, F., Groothoff, J., Van den Heuvel, W., & Post, D. (2000).
Reliability, validity and structure of the Adolescent Decision Making
Questionnaire among adolescents in The Netherlands. Personality and individual
differences, 28(2), 273-286.
Uranitsch, A. T. (2008). Autonom werden im Kontext der Familie.
von Sydow, K., Lieb, R., Pfister, H., Höfler, M., & Wittchen, H. U. (2002). What predicts
incident use of cannabis and progression to abuse and dependence?:: A 4-year
prospective examination of risk factors in a community sample of adolescents and
young adults. Drug and alcohol dependence, 68(1), 49-64.
133
Weller, N., Tortolero, S., Kelder, S., Grunbaum, J., Carvajal, S., & Gingiss, P. (1999).
Health risk behaviors of Texas students attending dropout prevention/recovery
schools in 1997. Journal of School Health, 69(1), 22-28.
White, H. R., Labouvie, E. W., & Papadaratsakis, V. (2005). Changes in Substance Use
During the Transition to Adulthood: A Comparison of College Students and Their
Noncollege Age Peers. Journal of Drug Issues.
Willett, J. B., & Sayer, A. G. (1994). Using covariance structure analysis to detect
correlates and predictors of individual change over time. Psychological Bulletin,
116(2), 363.
Williams, R. L. (2007). First-year university students' approach to their current
developmental period: ProQuest.
Williamson, M. D. (2007). An Exploration of the Relationships Between Blogging
Practices, Blogging Motives, and Identity Exploration.
Winters, K. (1990). The need for improved assessment of adolescent substance
involvement. Journal of Drug Issues.
YRBSS. (2009). Data and Statistics YRBSS: Youth Risk Behavior Surveillance System.
Zucker, R. A., Fitzgerald, H. E., & Moses, H. D. (1995). Emergence of alcohol problems
and the several alcoholisms: A developmental perspective on etiologic theory and
life course trajectory.
134
APPENDIX : MPLUS CODES
Model 1. Configural Invariance with 8 items.
TITLE: CONFIGURAL INVARIANCE;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\tnd2yr.dat ;
VARIABLE: NAMES = A1-A8 B1-B8 C1-C8 id cond;
USEVAR = A1-A8 B1-B8 C1-C8;
MISSING = .;
MODEL:
F1 by A1-A8;
F2 by B1-B8;
F3 by C1-C8;
OUTPUT: MODINDICES (ALL);
Model 2. Configural Invariance with 5 items.
TITLE: CONFIGURAL INVARIANCE;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\tnd2yr.dat ;
VARIABLE: NAMES = A1-A8 B1-B8 C1-C8 id cond;
USEVAR = A2-A3 A5-A7 B2-B3 B5-B7 C2-C3 C5-C7;
MISSING = .;
MODEL:
F1 by A2-A3;
F1 by A5-A7;
F2 by B2-B3;
F2 by B5-B7;
F3 by C2-C3;
F3 by C5-C7;
135
Model 3. Metric (weak) Invariance.
TITLE: METRIC INVARIANCE - LOADINGS + INTERCEPTS;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\tnd2yr.dat ;
VARIABLE: NAMES ARE A1-A8 B1-B8 C1-C8 SCHOOL PROGRAM;
USEVAR = A2-A3 A5-A7 B2-B3 B5-B7 C2-C3 C5-C7;
MISSING = .;
MODEL:
!Factor loadings invariant across waves
F1 by A2-A3(1-2);
F1 by A5-A7(3-5);
F2 by B2-B3(1-2);
F2 by B5-B7 (3-5);
F3 by C2-C3 (1-2);
F3 by C5-C7 (3-5);
OUTPUT: SAMP STAND RES TECH1 MODINDICES;
Model 4. Strong Invariance.
TITLE: METRIC INVARIANCE - LOADINGS + INTERCEPTS;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\tnd2yr.dat ;
VARIABLE: NAMES ARE A1-A8 B1-B8 C1-C8 SCHOOL PROGRAM;
USEVAR = A2-A3 A5-A7 B2-B3 B5-B7 C2-C3 C5-C7;
MISSING = .;
MODEL:
!Factor loadings invariant across waves
F1 by A2-A3(1-2);
F1 by A5-A7(3-5);
F2 by B2-B3(1-2);
F2 by B5-B7 (3-5);
F3 by C2-C3 (1-2);
F3 by C5-C7 (3-5);
!Intercepts invariant across waves
[A2 B2 C2] (10);
[A3 B3 C3] (11);
[A5 B5 C5] (13);
[A6 B6 C6] (14);
[A7 B7 C7] (15);
136
OUTPUT: SAMP STAND RES TECH1 MODINDICES;
Model 5. Strict Invariance.
TITLE: METRIC INVARIANCE - LOADINGS + INTERCEPTS;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\tnd2yr.dat ;
VARIABLE: NAMES ARE A1-A8 B1-B8 C1-C8 SCHOOL PROGRAM;
USEVAR = A2-A3 A5-A7 B2-B3 B5-B7 C2-C3 C5-C7;
MISSING = .;
MODEL:
!Factor loadings invariant across waves
F1 by A2-A3(1-2);
F1 by A5-A7(3-5);
F2 by B2-B3(1-2);
F2 by B5-B7 (3-5);
F3 by C2-C3 (1-2);
F3 by C5-C7 (3-5);
!Intercepts invariant across waves
[A2 B2 C2] (10);
[A3 B3 C3] (11);
[A5 B5 C5] (13);
[A6 B6 C6] (14);
[A7 B7 C7] (15);
!Variances invariant across waves
A2 B2 C2 (18);
A3 B3 C3 (19);
A5 B5 C5 (21);
A6 B6 C6 (22);
A7 B7 C7 (23);
OUTPUT: SAMP STAND RES TECH1;
137
Model 6. Linear growth model without covariates.
TITLE: Basic Growth Model;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\factor_tnd2yr.dat;
VARIABLE: NAMES ARE ID F1 F2 F3 SCHOOL COND AGE SEX;
USEVAR ARE F1 F2 F3;
MISSING ARE .;
MODEL:
i s | F1@0 F2@1 F3@2;
OUTPUT: SAMPSTAT TECH1 ;
PLOT: TYPE = PLOT3;
SERIES = F1-F3(*);
Model 7. Linear growth model with covariates.
TITLE: Basic Growth Model;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\factor_tnd2yr.dat;
VARIABLE: NAMES ARE ID F1 F2 F3 SCHOOL COND AGE SEX;
USEVAR ARE F1 F2 F3 COND AGE SEX;
MISSING ARE .;
MODEL:
i s | F1@0 F2@1 F3@2;
i s on COND AGE SEX;
OUTPUT: SAMPSTAT TECH1 ;
PLOT: TYPE = PLOT3;
SERIES = F1-F3(*);
138
Model 8. Latent class analysis with 3 classes.
TITLE: Latent Class w 3 classes;
DATA: FILE IS C:\Users\Nadra Lisha\Desktop\factor_tnd2yr.dat;
VARIABLE: NAMES ARE ID F1 F2 F3 SCHOOL COND AGE SEX;
USEVAR ARE F1 F2 F3 COND AGE SEX;
IDVARIABLE = ID;
MISSING ARE .;
CLASSES = C(3);
ANALYSIS: TYPE = MIXTURE;
STARTS =100 50;
STITERATIONS = 10;
MODEL:
%OVERALL%
i s | F1@0 F2@1 F3@2;
c i s on COND AGE SEX;
OUTPUT: SAMPSTAT TECH11 TECH14 ;
SAVEDATA:
file is C:\Users\Nadra Lisha\Desktop\EA_save.dat;
save is cprob;
format is free;
PLOT: TYPE IS PLOT3;
SERIES IS F1 - F3 (*);
Abstract (if available)
Abstract
The present dissertation project examined emerging adulthood in the context of substance use behavior. It is presently recognized that individuals pass through a transitional period between adolescence and young adulthood - “emerging adulthood.” Continuation high school youth have demonstrated to show early entry into emerging adulthood-type developmental tasks. The first study examined the psychometric properties of the Inventory of Dimensions of Emerging Adulthood (IDEA) in a population of continuation high school students of Southern California. The IDEA was developed in an attempt to capture the psychosocial attributes of this unique developmental period. A 21-item version of the IDEA was evaluated, and it was determined that the measure was composed of three dimensions and demonstrated high internal consistency. In addition, construct validity was assessed and indicated that the measure, as expected, was correlated with demographic characteristics, risk behaviors, and other psychological measures. It was concluded that the IDEA is a practical instrument for measuring emerging adulthood in at-risk populations. Not every individual experiences emerging in the same fashion
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Social self-control and adolescent substance use
PDF
Sociocultural stress, coping and substance use among Hispanic/Latino adolescents
PDF
Contextualizing social network influences on substance use among high risk adolescents
PDF
Role transitions, past life events, and their associations with multiple categories of substance use among emerging adults
PDF
The influence of contextual factors on the processes of adoption and implementation of evidence-based substance use prevention and tobacco cessation programs in schools
PDF
Motivational interviewing with adolescent substance users: a closer look
PDF
Exploring the role of peer influence, linguistic acculturation, and social networks in substance use
PDF
Post-traumatic growth among high-risk youth: predictors, impact of stressful life events, and relationship with changes in substance use behaviors
PDF
Distress tolerance and mindfulness disposition: associations with substance use during adolescence and emerging adulthood
PDF
Using a structural model of psychopathology to distinguish relations between shared and specific features of psychopathology, smoking, and underlying mechanisms
PDF
The Internet activities, gratifications, and health consequences related to compulsive Internet use
PDF
Factors and correlates of sexual behaviors among young adults from continuation high schools
PDF
Problematic alcohol use in Hispanic emerging adults: the role of perceived discrimination, cultural identity, and salient cultural values
PDF
Energy drink consumption, substance use and attention-deficit/hyperactivity disorder among adolescents
PDF
Intrapersonal and environmental factors associated with Chinese youth alcohol use experimentation and binge drinking behaviors
PDF
Prospective associations of stress, compulsive internet use, and posttraumatic growth among emerging adults
PDF
A network analysis of online and offline social influence processes in relation to adolescent smoking and alcohol use
PDF
A sociocultural and developmental approach to intimate partner violence among a sample of Hispanic emerging adults
PDF
Understanding the dynamic relationships between physical activity and affective states using real-time data capture techniques
PDF
The role of social support in the relationship between adverse childhood experiences and addictive behaviors across adolescence and young adulthood
Asset Metadata
Creator
Lisha, Nadra Erin
(author)
Core Title
The dynamic relationship of emerging adulthood and substance use
School
Keck School of Medicine
Degree
Doctor of Philosophy
Degree Program
Preventive Medicine (Health Behavior)
Publication Date
05/02/2012
Defense Date
03/02/2012
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
at-risk youth,continuation high school youth,growth curve modeling,latent growth analysis,OAI-PMH Harvest,substance use
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Sussman, Steven (
committee chair
), John, Richard S. (
committee member
), Leventhal, Adam M. (
committee member
), Rohrbach, Louise Ann (
committee member
), Sun, Ping (
committee member
)
Creator Email
nadralisha@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c3-24052
Unique identifier
UC11289398
Identifier
usctheses-c3-24052 (legacy record id)
Legacy Identifier
etd-LishaNadra-724.pdf
Dmrecord
24052
Document Type
Dissertation
Rights
Lisha, Nadra Erin
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 a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
at-risk youth
continuation high school youth
growth curve modeling
latent growth analysis
substance use