research paper-College student substance abuse

Peer Associations for Substance Use and Exercise in a College Student Social Network

Nancy P. Barnett, Miles Q. Ott, Michelle L. Rogers, Michelle Loxley, Crystal Linkletter, and Melissa A. Clark

Brown University

Substance use and exercise have opposite trajectories in young adulthood, and research indicates that peers are influential for both of these health behaviors, but simultaneous investigations of peer associ- ations with substance use and exercise have not been conducted. Objective: Use a college residence hall peer network to examine associations between peer behaviors and alcohol use, marijuana use, and exercise behavior. Method: 129 undergraduates (51.9% female, 48.1% non-Hispanic White; 84.5% first-year students) in one residence hall completed a Web-based survey of substance use and exercise and identified up to 10 students in the residence hall who were important to them. Two social network analytic methods, community detection cluster analysis and network autocorrelation modeling, were used to identify peer groupings and to examine the associations between peer and participant behaviors, respectively. Results: Participants nominated an average of 4.1 residence hall members, and 53.9% of the ties were reciprocal. 6 clusters were identified that differed significantly on demographics, college activities, substance use, and exercise. Weekly volume of alcohol consumed among nominated peers was significantly associated with that of participants, and all other covariates, including gender and athlete status, were not significant. Peer marijuana use also was associated with participant use after controlling for covariates. Exercise levels of nominated peers were not associated with exercise levels of participants. Conclusions: College student networks may be good targets for health-related prevention programs. Programs that use close-proximity peers to influence the behavior of others might be more effective with substance use as the target behavior than exercise.

Keywords: college, alcohol, exercise, social networks, peers

Supplemental materials: http://dx.doi.org/10.1037/a0034687.supp

The college years are a critical time for the development of positive and negative health behaviors that persist into later life. One of the most significant behavioral changes upon the transition to college is an increase in alcohol use (Borsari, Murphy, & Barnett, 2007), and one third of college marijuana users initiate use in college (Gledhill-Hoyt, Lee, Strote, & Wechsler, 2000). There is also evidence of a steep decline in physical activity through

adolescence and young adulthood (U.S. Department of Health & Human Services, 2008) and in 2012 at least 49% of over 90,000 college students surveyed did not meet national guidelines for exercise (American College Health Association, 2012). Given these shifts in healthy and unhealthy behaviors in late adolescence and the evidence that peers are influential for both, it is important to study their shared social context.

Peer Influences in Health Behaviors

Friendship groups tend to show similarities in behaviors and attitudes (Burk, van der Vorst, Kerr, & Stattin, 2012); this homo- geneity is theorized to be due to selection (sometimes called homophily; McPherson, Smith-Lovin, & Cook, 2001) and social- ization. Selection reflects the tendency of individuals to choose friends who are similar to themselves, and socialization is a pro- cess whereby individuals learn and adhere to norms of behavior within their social group (Steglich, Snijders, & Pearson, 2010). A robust literature supports the importance of these peer dynamics for understanding adolescent and young adult substance use (Bau- man & Ennett, 1994; Borsari & Carey, 2001) and exercise (Mackey & La Greca, 2007; Voorhees et al., 2005). For example, Demartini, Prince, and Carey (2013) found that the presence of a large number of heavy drinkers in a subcommunity of students was an important condition for the subsequent spread of heavy drinking in the group. In a study of peer networks in which students

This article was published Online First December 23, 2013. Nancy P. Barnett, Center for Alcohol and Addiction Studies and De-

partment of Behavioral and Social Sciences, Brown University; Miles Q. Ott, Department of Biostatistics, Brown University; Michelle L. Rogers, Center for Population Health and Clinical Epidemiology, Brown Univer- sity; Michelle Loxley, Center for Alcohol and Addiction Studies, Brown University; Crystal Linkletter, Department of Biostatistics, Brown Univer- sity; Melissa A. Clark, Center for Population Health and Clinical Epide- miology and Department of Epidemiology and Obstetrics and Gynecology, Brown University.

This research was supported by a Research Excellence Award from the Center for Alcohol and Addiction Studies, Brown University to Nancy Barnett. Miles Q. Ott is now at Carleton College and Crystal Linkletter is now at The MathWorks, Needham, MA.

Correspondence concerning this article should be addressed to Nancy P. Barnett, PhD, Brown University, Box G-S121-5, Providence, RI 02912. E-mail: [email protected]

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

Health Psychology © 2013 American Psychological Association 2014, Vol. 33, No. 10, 1134 –1142 0278-6133/14/$12.00 http://dx.doi.org/10.1037/a0034687

1134

reported on up to eight individuals in their network, Reifman, Watson, and McCourt (2006) found that having “drinking bud- dies” in one’s network at baseline predicted an increase in the respondent’s drinking later. Having a high proportion of fraternity/ sorority members in one’s network also was related to heavier drinking. This latter investigation is one of a few studies of college students that used network methods, but similar to other studies of peer influence on alcohol use (e.g., Wood, Read, Mitchell, & Brand, 2004), was limited in that it relied on the participant’s report of peer drinking.

In an investigation of young adult substance use in which peer reports were used, Andrews, Tildesley, Hops, and Li (2002) measured the alcohol use, cigarette use, and marijuana use of young adult participants and two peers. Peer reports of binge drinking and cigarette use were associated with partici- pant use, with a more limited effect for marijuana use. How- ever, there was no information about connections between surveyed individuals, precluding a more complete understand- ing of associations within the peer community. In a complete network of 34 members of one fraternity, Phua (2011) found a significant homophily effect for cigarette smoking but not for number of alcohol drinks per day. Similar to Demartini et al. (2013) this study also showed changes in drinking among network members according to who they “hung out” with. Therefore, despite the robust literature supporting the impor- tance of peer dynamics for substance use (see also Borsari & Carey, 2001; Burk et al., 2012), the research with college students is limited by (a) designs that do not allow for infer- ences about networks (Brechwald & Prinstein, 2011); and (b) indirect and typically biased measures of peer behavior derived from participant reports (Bourgeois & Bowen, 2001).

With regard to exercise, there is evidence that adolescents affiliate with peers who show similar levels of physical activity (de la Haye, Robins, Mohr, & Wilson, 2011; Voorhees et al., 2005). Affiliating with a particular crowd is also associated with exercise, with adolescents who affiliate with a “jock” or “popular” group exercising more than others (Mackey & La Greca, 2007). The literature supporting the importance of peer influence for exercise with college age participants is consis- tent, though smaller (e.g., Gruber, 2008). Important age group and gender differences in group affiliation and peer influences have been found. For example, Baker, Little, and Brownell (2003) found that the perception of peer behaviors was signif- icantly related to boys’ intention to exercise but not girls’, and Sallis, Taylor, Dowda, Freedson, and Pate (2002) found that older but not younger adolescents showed associations between peer support for exercise and their own vigorous exercise. This evidence suggests that individual differences are critical to measure and control for. In addition, more studies are needed that use direct reports from peers. In a recent review on peer influence in physical activity, Fitzgerald, Fitzgerald, and Aherne (2012) recommend using data collected directly from network members to improve ecological validity. Finally, de- termining whether subcommunities based on peer connections (i.e., ties) have similarities on important (desirable and unde- sirable) behaviors would be informative for prevention ap- proaches that target at-risk subgroups.

Social Network Analysis

Social network analysis (SNA) is a methodological and analytic approach that can be used to understand the risk and protective influences on individuals within a social context (Wasserman & Faust, 1994). SNA methods often entail exploring interpersonal connections within a shared location (e.g., a school). Such methods are particularly applicable to the current investigation because college students commonly have an intensely social living envi- ronment that influences their beliefs and behavior even in the first few months of school (Bourgeois & Bowen, 2001). One value of complete networks is that they provide relational data from which subcommunities of people and clusters of targeted behaviors can be appreciated. SNA has been used to understand the influence of adolescent peer networks on substance use (Burk et al., 2012; Ennett & Bauman, 1994) and exercise (de la Haye et al., 2011), and a few studies have investigated egocentric networks of young adults (Lau-Barraco & Collins, 2011), but rarely has a complete network been studied to understand substance use or exercise in college students.

Correlates of Health Behaviors

Several demographic characteristics and individual activities are associated with substance use and exercise. Notably, males, White students, members of fraternities/sororities, and athletes drink more heavily than other students (Buscemi, Martens, Murphy, Yurasek, & Smith, 2011; Wechsler, Dowdall, Daven- port, & Castillo, 1995). Adolescent and young adult males smoke more marijuana than females, White students smoke more marijuana than Hispanic and Black students, and boys and White individuals exercise more than girls and non-Whites, respectively (Buscemi et al., 2011; Johnston, O’Malley, Bach- man, & Schulenberg, 2012). Furthermore, there is evidence that alcohol-related problems are higher in the first-year of college (Harford, Wechsler, & Muthen, 2003), and class-year differ- ences in marijuana use have been reported (Gledhill-Hoyt et al., 2000). Research on the association between peer behaviors and college student substance use and exercise must control for these other relationships.

The objective of this study was to investigate the association of peer behavior with alcohol use, marijuana use, and exercise in one residence hall (RH) of college students, with a focus on first-year students. A RH network was investigated because many first-year college relationships will be among members in the same RH, and because a RH provides a reasonably sized closed population that can be fully observed. Two methods for investigating peer asso- ciations were used. First, subcommunities of participants were identified based solely on peer nominations to determine the extent to which substance use and exercise were clustered in these sub- groups. A community detection approach (cluster analysis) iden- tified more tightly knit groups, and comparisons of these groups investigated the clustering of demographics, activities, and health behaviors. Next, the behaviors of nominated peers were investi- gated as correlates of alcohol, marijuana use, and exercise in participants, controlling for demographics and campus-based ac- tivities commonly associated with substance use and/or exercise.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1135PEER ASSOCIATIONS FOR SUBSTANCE USE AND EXERCISE

Method

Participants

One RH on the campus of a university in the Northeast was selected for its moderate size. Students (N � 194) were identified using a resident list provided by the university. Participants who were not 18 (n � 6) were excluded, resulting in 188 eligible students. Of age-eligible students, 129 consented (68.6% response rate), and all completed the survey; five (2.7%) declined to par- ticipate but allowed their names to remain on the nomination list, and nine students (4.8%) declined to participate and “opted-out” of having their names on the nominations list. The remaining 45 participants did not respond.

Procedures

Near the end of the fall semester a “look for our survey” invitation letter and e-mail and a $5 gift card were mailed to the RH residents explaining the study. The next week, an invitation containing a link to the Web-based survey was sent by e-mail. The first page of the survey contained the consent text. Students had the option to: (a) enroll in the study; (b) not enroll in the study but allow their name to be retained on the network list; or (c) not enroll in the study and have their name removed from the network list (i.e., “opt-out”). Regular reminder e-mails were sent to nonre- sponders. The survey was available for 5 weeks and was discon- tinued when students left campus for the winter break. Participants were compensated $20 and procedures were approved by the university Institutional Review Board.

Measures

Demographics. Respondents were asked to provide gender and year in school. Race was collected separately from ethnicity and respondents could endorse more than one race.

Activities. Respondents were asked, “Are you a member of a sorority or fraternity?” Those who answered “No” were asked, “Are you planning to become a member of a sorority or frater- nity?” Answer choices for this question were “No,” “Yes,” and “Maybe.” Participants answering “Yes” or “Maybe” were coded as considering joining a fraternity/sorority. Athlete status was mea- sured with the question, “Are you a member of a varsity athletic team?”

Substance use and exercise. Alcohol use was measured using the Graduated Frequency for Alcohol Questionnaire (Hilton, 1989) from which number of drinks per week in the past semester was derived. Total number of alcohol-related problems in the past semester was measured with the 24-item Brief-Young Adult Al- cohol Consequences Questionnaire (Cronbach’s alpha � .87; Kahler, Strong, & Read, 2005). Marijuana use was measured by asking, “Have you used marijuana since the start of the school year?” Exercise activity was measured as the number of minutes per day in the past week of moderate (“activities that take mod- erate physical effort and make you breathe somewhat harder than normal”) and vigorous (“activities that take hard physical effort and make you breathe much harder than normal”) exercise (Booth, 2000).

Social network. The social network measure asked respon- dents to enter the first name and last initial of up to 10 people “who have been important to you since the start of the school year. These might be people you socialized with, studied with, or regularly had fun with. These people may be roommates, friends, family mem- bers, people from work, or anyone that you see as having had a significant impact on your life, regardless of whether or not you liked them” (adapted from the Important People Instrument; Long- abaugh & Zywiak, 2002). Participants then indicated which of these individuals lived in their RH by clicking a box next to the name they had entered. A calculation within the survey summed the number of nominated peers who lived in the RH and invited the participant to nominate additional RH peers, up to 10.

To create the network structure, participants were asked to select their RH peers’ names from a prepopulated dropdown list of all RH residents. Students who had opted out of the study were not available for selection. For participants who wanted to nominate a peer who had opted out, the option “I cannot find this person in the list” was available. Any data collected about an individual before he or she opted out were recoded to missing. Students who did not respond to the survey invitation remained on the list for selection by participants but were not included in any network analyses because no self-reported behavior was available.

For descriptive purposes the number of nominations of partici- pating peers in the RH network and the proportion of reciprocated ties were calculated. The number of separate network components (portions of the network where there is a path between members within a component but no paths between the components), was identified using the SNA package in R version 2.1 (Butts, 2010; R Development Core Team, 2011).

Data Analysis

The final dataset contained de-identified data in which each participant’s self-report was coded by ID and linked to other participants’ self-reports, according to their nominations of each other. Information about out-of-network members was not ana- lyzed for this study. Participant descriptive information was ana- lyzed using independent t tests and chi-square tests (SPSS Statis- tics, IBM, Version 19). The drinks per week and alcohol problems variables were log transformed for analyses due to kurtosis in the distributions.

Cluster analysis. Clusters were identified using the Girvan- Newman edge-betweenness community detection method (New- man & Girvan, 2004) in the igraph package (Csardi & Nepusz, 2006) using the R statistical programming language (R Develop- ment Core Team, 2011). The Girvan-Newman method assesses which edges, or links between individuals, have the largest be- tweenness. An edge has high edge betweenness if it is included in the shortest path between multiple individuals in the network. Edges with high betweenness tend to connect otherwise uncon- nected communities or clusters. By progressively deleting the edges with the highest edge betweenness, clusters are identified. To test whether alcohol use, alcohol-related problems, and exer- cise varied by cluster, analysis of variance (ANOVA), and Tukey’s multiple comparison of means (Miller, 1981) were used to main- tain the proper level of Type I error. Chi-square tests determined if the binary measures of marijuana use, demographics (gender, race/ethnicity, class year), and activity variables (fraternity/soror-

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1136 BARNETT ET AL.

ity interest and athletic status) varied by cluster, with a post hoc comparison of proportions, using a Bonferrroni correction to ac- count for multiple comparisons.

Network analysis. To examine the relevance of peer associ- ations for behaviors measured on a continuous scale (drinks per week, alcohol-related problems, and exercise hours per week), a network autocorrelation model in R’s spdep package (Bivand, 2011) was used. Although a regular linear regression assumes independence of observations, a network autocorrelation model both allows for and attempts to estimate the correlation between people who are “neighbors” in the network (i.e., A is B’s neighbor if B nominated A or A nominated B). If the estimate of the correlation between neighbors on the behavior of interest is sig- nificantly different from zero, one can conclude there is a signif- icant network effect (i.e., peer behavioral association). Separate models investigated drinks per week, alcohol-related problems, and exercise. Additional variables included in each model were gender, race/ethnicity, class year, athlete status, and interest in joining a fraternity/sorority. For the binary variable of marijuana use in the past semester an analogous Bayesian probit network auto-correlation model using MATLAB (LeSage, 2010) was em- ployed. The same covariates above were included in this model, and the 95% credible interval1 of the network auto-correlation parameter was used to conclude whether or not the marijuana use of RH participants was significantly associated with their nomi- nated peers’ marijuana use.

Results

Sample Description

Demographics. Of the 129 participants, 67 (51.9%) were female; 109 (84.5%) were first-year students, 16 (12.4%) were sophomores, and four (3.1%) were juniors. Class year was subse- quently dichotomized as first year versus other. For race, 65 (50.4%) identified as White, 31 (24.0%) as Asian, 6 (4.7%) as Black, two (1.6%) as other races, 11 (8.5%) as more than one race, and 14 (10.9%) did not indicate a race. Hispanic/Latino ethnicity was reported by 14 (10.9%) participants. Due to small group sizes and because White race shows greater alcohol and drug-related risk (Johnston et al., 2012), race/ethnicity was dichotomized as non-Hispanic White (N � 62; 48.1%) or other (N � 63; 48.8%), with four participants missing this variable. In the sample, 35 (27.1%), were varsity athletes, with two individuals missing this item. No participants were members of a fraternity or sorority, but 31 (24.0%) were considering joining.

Alcohol use. The average number of drinks per week among participants was 8.3 (SD � 11.3). The average number of alcohol problems in the past semester was 3.3 (SD � 3.8). There was a trend for men to drink a higher number of drinks per week (10.2) than women (6.5), t(127) � 1.90, p � .059, but there was no gender difference on alcohol problems (3.1 for men, 3.4 for women). There was no difference by race/ethnicity in drinks per week (9.4 for non-Hispanic White students, 6.8 for other students), but non-Hispanic White students showed a significantly higher number of alcohol-related problems (3.9 vs. 2.4), t(123) � 2.29, p � .024. Individuals who were considering joining a fraternity or sorority were not significantly different in drinks per week com- pared with participants not considering joining, but there was a

trend for alcohol problems, t(127) � 1.98, p � .050, with those considering joining showing higher scores (4.4 vs. 2.9). Athletes reported a higher number of drinks per week (9.3) than nonathletes (7.8), t(125) � 2.02, p � .046, but there was no difference on alcohol problems by athlete status. For class year (dichotomized), there were not significant differences on the two alcohol variables.

Marijuana use. In the sample, 37 (29.8%) participants had used marijuana in the semester.2 There were no demographic differences on marijuana use. Students who were considering joining a fraternity or sorority showed a trend toward greater use of marijuana (43.3% vs. 25.5%), �2(1, N � 124) � 3.44, p � .064.

Exercise. Participants engaged in moderate exercise an aver- age of 2.1 hr per week (SD � 3.1) and in vigorous exercise an average of 5.7 hr per week (SD � 6.4). When moderate and vigorous exercise were combined, men reported more exercise per week than women (9.9 hr vs. 5.8 hr), t(127) � 3.13, p � .002, and non-Hispanic White students reported more exercise than other students (9.6 hr vs. 6.3 hr), t(123) � 2.44, p � .016. Not surpris- ingly, students on athletic teams engaged in more exercise than nonathletes (16.2 hr vs. 4.5 hr), t(125) � 8.78, p � .001. There were no significant class year or fraternity/sorority interest differ- ences in exercise.

Description of Network

Of the 129 participants who completed the survey, four (3.1%) did not nominate anyone in the network and were not nominated by anyone. These four isolates were excluded from the analyses because both the cluster and network analyses use ties to investi- gate peer associations. The 125 remaining participants nominated an average of 4.1 peers who lived in the RH who themselves participated in the study (SD � 2.8; Median � 4; range � 0 –10), totaling 507 nominations. Of the 507 nominations in the network, 272 (53.9%) were reciprocated. The 125 participants comprised two components (i.e., sections not connected by any ties).3 The larger component was composed of 123 participants, and a second component was composed solely of a dyad, where both partici- pants exclusively nominated the other.

Cluster Analysis

To avoid reporting identifying information on the dyad compo- nent (which would serve as its own cluster), these two participants were excluded from the cluster analysis, leaving one connected component of 123 students. Using the Girvan-Newman commu- nity detection method and inspection of a dendogram by the authors, six clusters were identified. Descriptive information about the clusters is presented in Table 1, and the network and clusters are presented in Figure 1. There were significant overall differ- ences among clusters on all characteristics except interest in joining a fraternity/sorority. Drinks per week (log transformed)

1 A Bayesian credible interval is analogous to a confidence interval, with a somewhat different interpretation: a 95% Bayesian credible interval is interpreted as there is a 95% chance that the true observed value is in this interval.

2 Five participants (3.9%) did not answer the marijuana use questions. 3 The total number of original components was six because the four

isolates discussed previously should also be considered components in the original sample.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1137PEER ASSOCIATIONS FOR SUBSTANCE USE AND EXERCISE

varied significantly by cluster, and using Tukey’s comparison tests, the largest cluster (Cluster 1) and two smallest clusters (Clusters 5 and 6) showed higher levels than the other three clusters. The (log transformed) alcohol problems score also varied significantly by cluster. Cluster 6 had significantly higher scores than Clusters 3 and 4, and Clusters 1, 2, and 5 had significantly higher scores than Cluster 3. Clusters 1 and 5 had significantly higher past-semester marijuana use than Cluster 3.4 Number of hours of exercise per week was higher in Clusters 1 and 5 than Cluster 3. Additional figures of the distribution of alcohol use, marijuana use, and exercise in the network are in supplemental materials.

Network Autocorrelation Models

Alcohol. Table 2 shows the results of the network autocorre- lation models for average number of drinks per week. The signif- icant (p � .001) lambda, which is the network autocorrelation parameter in the models for continuous outcomes, indicates that the drinking volume of nominated peers was significantly posi- tively associated with participant drinking volume, after control- ling for other variables. The nonsignificant values of the correlates indicate that peer drinking among participants’ nominated peers was the only significant variable associated with participant drink- ing.

Alcohol problems. Females, those interested in joining a fra- ternity/sorority, and heavier drinkers showed more alcohol-related problems, but the network autocorrelation lambda value for alco- hol problems in nominated peers was not significant, indicating that peer alcohol problems were not associated with a higher risk of problems for an individual after that individual’s characteristics were taken into consideration (see Table 2).

Marijuana. For past-semester marijuana use the network au- tocorrelation parameter’s 95% credible interval did not include zero, therefore, there was evidence of a significant positive net- work effect of past-semester marijuana use (i.e., use of nominated peers), after adjusting for covariates. Other than the network autocorrelation, only membership on an athletic team was signif- icantly (negatively) associated with marijuana use.

Exercise. There was no evidence of residual network autocor- relation for exercise, thus the amount of exercise among nominated peers was not associated with participant exercise (see Table 2). Compared with males, females reported less exercise, and being on an athletic team was associated with more exercise.

Discussion

Two different network analytic methods were used to investi- gate whether subcommunities of college students living together in one residence hall network showed similarities on substance use and exercise, and whether the behaviors of nominated peers were

T ab

le 1

S a m

p le

C h a ra

ct er

is ti

cs a n d

H ea

lt h

B eh

a vi

o rs

b y

C lu

st er

T ot

al sa

m pl

e C

lu st

er 1

C lu

st er

2 C

lu st

er 3

C lu

st er

4 C

lu st

er 5

C lu

st er

6 S

ta ti

st ic

(N �

12 3)

(N �

47 )

(N �

39 )

(N �

20 )

(N �

8) (N

� 5)

(N �

4)

P ro

po rt

io n

m al

e 0.

47 0.

47 a

0. 38

a 0.

35 a

0. 75

a 1.

00 a

1. 00

a �

2 (d

f �

5) �

14 .8

9, p

� .0

11 P

ro po

rt io

n no

n- H

is pa

ni c

W hi

te 0.

49 0.

55 a

0. 56

a 0.

10 b

0. 75

a 0.

80 a ,b

0. 00

a ,b

� 2 (d

f �

5) �

21 .7

1, p

� .0

01 P

ro po

rt io

n fi

rs t-

ye ar

st ud

en ts

† 0.

88 —

— —

— —

— �

2 (d

f �

5) �

70 .3

8, p

� .0

01 P

ro po

rt io

n co

ns id

er in

g jo

in in

g a

fr at

er ni

ty /s

or or

it y

0. 24

0. 34

0. 23

0. 15

0. 13

0. 00

0. 00

� 2 (d

f �

5) �

7. 00

, p

� .2

21 P

ro po

rt io

n at

hl et

es ‡

0. 28

0. 45

a 0.

18 a

0. 10

a 0.

00 a

0. 80

a 0.

00 a

� 2 (d

f �

5) �

23 .2

0, p

� .0

01 M

ea n

nu m

be r

of dr

in ks

pe r

w ee

k (S

D )�

8. 62

(1 1.

50 )

12 .8

0 (1

2. 43

)a 6.

34 (8

.2 5)

b 1.

53 (4

.6 6)

b 0.

38 (0

.6 2)

b 20

.2 3

(2 0.

26 )a

19 .1

0 (1

5. 09

)a F

(5 ,

11 7)

� 12

.1 6,

p �

.0 01

M ea

n nu

m be

r of

al co

ho l

pr ob

le m

s in

pa st

se m

es te

r (S

D )�

3. 33

(3 .8

4) 4.

64 (3

.7 3)

a ,c

3. 10

(3 .9

4) a ,c

0. 60

(1 .7

0) b

0. 75

(1 .3

9) a ,b

4. 80

(5 .0

7) a ,c

7. 25

(3 .5

0) c

F (5

, 11

7) �

9. 61

, p

� .0

01 P

ro po

rt io

n re

po rt

in g

pa st

se m

es te

r m

ar ij

ua na

us e�

� 0.

31 0.

48 a

0. 30

a ,b

0. 05

b 0.

00 a ,b

0. 25

a ,b

0. 67

a �

2 (d

f �

5) �

17 .7

6, p

� .0

01 M

ea n

nu m

be r

of ex

er ci

se ho

ur s

in pa

st w

ee k

(S D

) 7.

95 (7

.7 8)

10 .5

7 (7

.7 7)

a 6.

38 (6

.9 3)

a ,b

4. 25

(5 .9

0) b

4. 68

(7 .4

5) a ,b

15 .5

0 (1

2. 13

)a 8.

00 (5

.2 5)

a ,b

F (5

, 11

7) �

3. 94

, p

� .0

02

N o te

. V

al ue

s w

it h

di ff

er en

t su

pe rs

cr ip

ts in

di ca

te st

at is

ti ca

ll y

si gn

if ic

an t

pa ir

w is

e di

ff er

en ce

s us

in g

a B

on fe

rr on

i co

rr ec

ti on

(� �

0. 05

/1 5

� 0.

00 3)

fo r

m ul

ti pl

e co

m pa

ri so

ns fo

r bi

na ry

ou tc

om es

, an

d T

uk ey

’s m

ul ti

pl e

co m

pa ri

so n

of m

ea ns

fo r

co nt

in uo

us ou

tc om

es .

† In

or de

r to

li m

it id

en ti

fi ab

il it

y on

ly th

e ov

er al

l pr

op or

ti on

of fi

rs t-

ye ar

st ud

en ts

ar e

pr es

en te

d, no

t th

e cl

us te

r- sp

ec if

ic pr

op or

ti on

s of

fi rs

t- ye

ar st

ud en

ts .

‡ A

ss um

in g

th at

tw o

in di

vi du

al s

(o ne

in C

lu st

er 1

an d

on e

in C

lu st

er 2)

w ho

pr ef

er re

d no

t to

an sw

er ar

e no

t at

hl et

es .

� S

ta ti

st ic

al te

st s

pe rf

or m

ed on

lo g-

tr an

sf or

m ed

va ri

ab le

, un

tr an

sf or

m ed

va lu

es ar

e pr

es en

te d

in ta

bl e

fo r

ea se

of in

te rp

re ta

ti on

. � �

E xc

lu di

ng th

e fi

ve in

di vi

du al

s w

ho pr

ef er

re d

no t

to an

sw er

(o ne

in C

lu st

er 1,

tw o

in C

lu st

er 2,

on e

in C

lu st

er 5,

an d

on e

in C

lu st

er 6)

.

4 Five participants did not answer the past-semester marijuana question. A sensitivity analysis was conducted in which participants who did not answer were first treated as past-semester marijuana users and then as nonusers. For these analyses, the chi-square values also were statistically significant (�2 � 19.33, �2 � 16.74, respectively), ps � .05, indicating that the missing data did not affect the findings.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1138 BARNETT ET AL.

related to participant behaviors. The community detection cluster analysis used only directed ties to detect subcommunities of indi- viduals, and the comparison of those groups established that they differed significantly on demographic, activity, and behavior pro- files. The two largest clusters (1 and 2) differed primarily on the amount of alcohol use, with Cluster 1 showing twice as many drinks per week as Cluster 2. The next largest cluster (3) was comprised of mostly non-White students and had the lowest num- ber of alcohol problems and very low rates of marijuana use. Clusters 4 and 5 were mostly male and non-Hispanic White but were on opposite ends of the spectrum for alcohol use and exer- cise; Cluster 5 had a high proportion of athletes, the highest exercise, and high drinking rates whereas Cluster 4 had no athletes, low exercise, and the lowest drinking rates. The smallest cluster

(6) had high alcohol use, the highest number of alcohol problems, and the highest marijuana use. Next, using network autocorrelation modeling and controlling for demographic and activity variables, the weekly volume of alcohol consumed by close peers (i.e., those nominated by the participant) was the primary correlate of a participant’s weekly alcohol volume. The importance of peer al- cohol use for college student drinking is not a novel finding per se (Borsari & Carey, 2001, 2006), but the nonsignificance of gender, race/ethnicity, class year, interest in joining a fraternity/sorority, and athlete status in the model indicates that peer drinking may be more relevant to college student drinking than personal and activ- ity characteristics. Furthermore, many previous investigations re- lied on student perceptions of peer behavior or global measures such as “How much do your friends drink?” whereas the current

Figure 1. Six clusters derived from network nominations (N � 123). Arrows depict the direction of the nomination, with the origin of the arrow showing the nominator and bidirectional arrows indicating reciprocated nominations.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1139PEER ASSOCIATIONS FOR SUBSTANCE USE AND EXERCISE

study used actual peer reports, thereby improving confidence in the accuracy of data about the peers.

Peer alcohol problems were not associated with participant alcohol problems, and the strongest correlate of participant alcohol problems was participant alcohol consumption. Taking the net- work findings on alcohol use and problems together, one can conclude that peer drinking is associated with participant drinking, but that participant drinking (but not peer problems) likely drives the experience of participant problems. This finding is consistent with research that found that the clustering of nonviolent alcohol consequences in communities was nonsignificant after the alcohol use of the participants was included in the analysis (Reboussin, Song, & Wolfson, 2012).

Similar to the findings for alcohol use, the marijuana use of nominated peers was significantly associated with past-semester marijuana use in participants. The only individual difference vari- able that was associated with marijuana use was athlete status, with athletes less likely to report marijuana use. Athletes may smoke marijuana less due to concern about the effects of smoking on athletic performance and/or to concern about drug testing. Never- theless, the findings for alcohol use and marijuana use are similar, with the substance use of nominated peers being a primary corre- late of participant substance use.

Contrary to the findings for alcohol use and marijuana use, participant exercise behavior was not significantly associated with peer exercise. Whereas peer support may be positively related to youth exercise (Fitzgerald et al., 2012), our findings suggest that the actual amount of exercise engaged in by close peers may not be as important as other characteristics. This conclusion is tentative because this study had a small sample size and was cross-sectional, but it nonetheless contributes a possible new perspective to the research literature on the importance of peers in the participation of young adults in different health behaviors.

Findings from this investigation demonstrate that peer clusters are differentiated by the behaviors of individuals in those clusters, but when relevant correlates of behavior are included in analyses, there are important differences that emerge with regard to risk and protective behaviors. A conclusion that could be drawn is that for young adults the peer dynamics of homophily may be stronger for substance use than for exercise. This difference may be due to motivational or contextual differences in alcohol use (most often done socially; Christiansen, Vik, & Jarchow, 2002) and exercise (perhaps more commonly done alone), possibilities that warrant additional investigation. Findings are consistent with other studies that find that adolescent peer behavior is more influential for problem behaviors than for protective behavior (Aalsma, Carpen- tier, Azzouz, & Fortenberry, 2012), and that perceptions of peer behaviors are more closely associated with intentions to engage in health risk-behaviors than health-promoting behaviors (Rivis & Sheeran, 2003). It is also possible that in the context of deviant norms for alcohol and marijuana use on college campuses, the more prosocial or protective behaviors are less likely to be social- ized (Oetting & Donnermeyer, 1998).

The enrollment rate for this investigation was high for survey studies but fell just below the traditionally accepted rate of 70% for network studies. Some students opted out or did not respond, and students under the age of 18 were excluded. Whether enrolled participants differed from nonresponders cannot be determined. The missing network data from nonresponders could have affected findings, but the alcohol use values were consistent with other investigations from the same campus (e.g., Hoeppner et al., 2012). This investigation did not include one dyad and several isolates; the isolates in particular are important to study because they may be at higher risk for substance use (Ennett et al., 2006). Measures of substance use covered the past semester but the measure of exercise covered the past week; it is possible that if measured over

Table 2 Correlates of Drinks per Week, Alcohol-Related Problems, Marijuana Use, and Hours of Exercise (N � 125)

Drinks per week Alcohol problemsa Marijuana useb Exercise hours

Correlates Beta SE z value Beta SE z value Betac 95% CI Beta SE z value

Intercept 1.62 0.27 5.97��� �0.13 0.13 �1.05 �0.20 [�0.72, 0.33] 371.85 61.38 6.06���

Gender Male Reference Reference Reference Reference Female �0.12 0.21 �0.59 0.36 0.1 3.59��� �0.08 [�0.59, 0.42] �140.56 58.12 �2.42�

Race/ethnicity Non-Hispanic White Reference Reference Reference Reference Other �0.22 0.18 �1.19 0.01 0.1 0.07 �0.34 [�0.86, 0.18] �64.72 50.55 �1.07

College year First year Reference Reference Reference Reference Other 0.22 0.36 0.62 0.13 0.15 0.91 0.01 [�0.75, 0.72] 96.40 80.32 1.20

Considering joining fraternity/ sorority 0.26 0.20 1.27 0.27 0.11 2.38� 0.51 [�0.06, 1.09] 34.87 69.23 0.50

Athlete �0.18 0.24 �0.77 0.06 0.11 0.53 �0.62 [�1.22, �0.06] 674.70 65.84 10.25���

drinks per week — — — 0.56 0.04 14.46��� — — — — — �d 0.60, LR � 33.91��� 0.11, LR � .70 0.32 [0.05, 0.59] �0.17, LR � 1.66

Note. The drinks per week and alcohol problems variables were log transformed. For drinks per week, alcohol problems, and exercise models are network autocorrelation models. For marijuana use the model is Bayesian Probit Network autocorrelation and the coefficients are probit regression coefficients. LR � likelihood ratio. a Measured with the Brief Young Adult Alcohol Consequences Questionnaire (B-YAACQ). b Missing reports imputed as no past-semester marijuana use. Results not different when missing reports imputed as having used marijuana. c Posterior Mean Value after a burn-in of 5,000 simulations and keeping 5,000 simulations. d Lambda is the network autocorrelation parameter, which reflects the importance of peer behavior in the participant’s individual network. � p � .05. ��� p � .001.

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1140 BARNETT ET AL.

a longer time frame, participant exercise would show greater associations with peer behavior. Connections outside the RH were not included in analyses, so findings do not reflect fully the personal networks of participants. Only one RH was used, the majority of students were first-year students, and race/ethnicity was dichotomized; larger networks are needed to understand how residence and demographics may interact with peer connections to influence behavior. Unmeasured variables, including exercise and substance use prior to college, and characteristics of the sample (e.g., the inclusion of student athletes) might have influenced results. Findings may not generalize to the campus at large, other class years, or other campuses. Finally, this investigation was cross-sectional so conclusions cannot be made about whether selection or socialization dynamics were more important; evidence of network associations was revealed but not the direction of these effects.

College students commonly reside on campuses that provide identifiable, accessible networks of individuals that may be good targets for investigation and for health-related prevention pro- grams. The importance of network effects during the transition to college and across changes in group affiliations and academic demands during college has been minimally investigated. For both substance use and protective behaviors, groupings in residence halls or other living units are an important target for investigation, as are student affiliations on shared interests such as athletics, fraternities/sororities, clubs, or academic concentrations. More- over, there is evidence that substance use and exercise have related trajectories in young adulthood, with increases in exercise being associated with decreased frequency of substance use (Terry- McElrath & O’Malley, 2011); therefore, it is important to simul- taneously investigate peer influences on these possibly competing behaviors. Investigating networks longitudinally would provide more information about how risk and protective behaviors are transmitted across networks over time, and investigating specific nominations (e.g., best friend) and shared activities (e.g., drinking events) and their importance to participants would be valuable. Finally, specific constructs that may be important for the trans- mission of peer influence, including social norms and social cap- ital, warrant further investigation.

Information about social networks on college campuses may be useful in planning prevention programs because social connections have the potential to transmit adaptive or positive health behaviors and/or change maladaptive behaviors. Indeed there is considerable evidence that peer networks can be supportive in substance use behavior change (Valente et al., 2007) and thus are a potential way prevention messages might be transmitted to those at risk. Findings from this study suggest that prevention programs that use close peers to influence the behavior of others might be more effective with substance use as the target behavior than exercise, but further investigation of the potential of social networks for prevention efforts is needed.

References

Aalsma, M. C., Carpentier, M. Y., Azzouz, F., & Fortenberry, J. D. (2012). Longitudinal effects of health-harming and health-protective behaviors within adolescent romantic dyads. Social Science & Medicine, 74, 1444 –1451. doi:10.1016/j.socscimed.2012.01.014

American College Health Association. (2012). National College Health Assessment II: Reference Group Executive Summary Hanover, MD: American College Health Association.

Andrews, J. A., Tildesley, E., Hops, H., & Li, F. (2002). The influence of peers on young adult substance use. Health Psychology, 21, 349 –357. doi:10.1037/0278-6133.21.4.349

Baker, C. W., Little, T. D., & Brownell, K. D. (2003). Predicting adoles- cent eating and activity behaviors: The role of social norms and personal agency. Health Psychology, 22, 189 –198. doi:10.1037/0278-6133.22.2 .189

Bauman, K. E., & Ennett, S. T. (1994). Peer influence on adolescent drug use. American Psychologist, 49, 820 – 822. doi:10.1037/0003-066X.49 .9.820

Bivand, R. (2011). spdep: Spatial dependence: Weighting schemes, statis- tics and models. R package version 0.5–31.

Booth, M. L. (2000). Assessment of physical activity: An international perspective. Research Quarterly for Exercise and Sport, 71, s114 –s120.

Borsari, B., & Carey, K. B. (2001). Peer influences on college drinking: A review of the research. Journal of Substance Abuse, 13, 391– 424. doi:10.1016/S0899-3289(01)00098-0

Borsari, B., & Carey, K. B. (2006). How the quality of peer relationships influences college alcohol use. Drug and Alcohol Review, 25, 361–370. doi:10.1080/09595230600741339

Borsari, B., Murphy, J. G., & Barnett, N. P. (2007). Predictors of alcohol use during the first year of college: Implications for prevention. Addic- tive Behaviors, 32, 2062–2086. doi:10.1016/j.addbeh.2007.01.017

Bourgeois, M. J., & Bowen, A. (2001). Self-organization of alcohol-related attitudes and beliefs in a campus housing complex: An initial investi- gation. Health Psychology, 20, 434 – 437. doi:10.1037/0278-6133.20.6 .434

Brechwald, W. A., & Prinstein, M. J. (2011). Beyond homophily: A decade of advances in understanding peer influence processes. Journal of Re- search on Adolescence, 21, 166 –179. doi:10.1111/j.1532-7795.2010 .00721.x

Burk, W. J., van der Vorst, H., Kerr, M., & Stattin, H. (2012). Alcohol use and friendship dynamics: Selection and Socialization in early-, middle-, and late-adolescent peer networks. Journal of Studies on Alcohol and Drugs, 73, 89 –98.

Buscemi, J., Martens, M. P., Murphy, J. G., Yurasek, A. M., & Smith, A. E. (2011). Moderators of the relationship between physical activity and alcohol consumption in college students. Journal of American College Health, 59, 503–509. doi:10.1080/07448481.2010.518326

Butts, C. T. (2010). sna: Tools for Social Network Analysis. R package version 2.2– 0.

Christiansen, M., Vik, P. W., & Jarchow, A. (2002). College student heavy drinking in social contexts versus alone. Addictive Behaviors, 27, 393– 404. doi:10.1016/S0306-4603(01)00180-0

Csardi, G., & Nepusz, T. (2006). The igraph software package for complex network research. InterJournal, Complex Systems 1695.

de la Haye, K., Robins, G., Mohr, P., & Wilson, C. (2011). How physical activity shapes, and is shaped by, adolescent friendships. Social Science & Medicine, 73, 719 –728. doi:10.1016/j.socscimed.2011.06.023

Demartini, K. S., Prince, M. A., & Carey, K. B. (2013). Identification of trajectories of social network composition change and the relationship to alcohol consumption and norms. Drug and Alcohol Dependence, 132, 309 –315.

Ennett, S. T., & Bauman, K. E. (1994). The contribution of influence and selection to adolescent peer group homogeneity: The case of adolescent cigarette smoking. Journal of Personality and Social Psychology, 67, 653– 663. doi:10.1037/0022-3514.67.4.653

Ennett, S. T., Bauman, K. E., Hussong, A., Faris, R., Foshee, V. A., Cai, L., & DuRant, R. H. (2006). The peer context of adolescent substance use: Findings from social network analysis. Journal of Research on Adolescence, 16, 159 –186. doi:10.1111/j.1532-7795.2006.00127.x

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1141PEER ASSOCIATIONS FOR SUBSTANCE USE AND EXERCISE

Fitzgerald, A., Fitzgerald, N., & Aherne, C. (2012). Do peers matter? A review of peer and/or friends’ influence on physical activity among American adolescents. Journal of Adolescence, 35, 941–958. doi: 10.1016/j.adolescence.2012.01.002

Gledhill-Hoyt, J., Lee, H., Strote, J., & Wechsler, H. (2000). Increased use of marijuana and other illicit drugs at US colleges in the 1990s: Results of three national surveys. Addiction, 95, 1655–1667. doi:10.1046/j.1360- 0443.2000.951116556.x

Gruber, K. J. (2008). Social support for exercise and dietary habits among college students. Adolescence, 43, 557–575.

Harford, T. C., Wechsler, H., & Muthen, B. O. (2003). Alcohol-related aggression and drinking at off-campus parties and bars: A national study of current drinkers in college. Journal of Studies on Alcohol, 64, 704 – 711.

Hilton, M. E. (1989). A comparison of a prospective diary and two summary recall techniques for recording alcohol consumption. British Journal of Addiction, 84, 1085–1092. doi:10.1111/j.1360-0443.1989 .tb00792.x

Hoeppner, B. B., Barnett, N. P., Jackson, K. M., Colby, S. M., Kahler, C. W., Monti, P. M., . . . Fingeret, A. (2012). Daily college student drinking patterns across the first year of college. Journal of Studies on Alcohol and Drugs, 73, 613– 624.

IBM SPSS for Windows Version 19.0 [Computer Software] (2010). New York: IBM Corp.

Johnston, L. D., O’Malley, P. M., Bachman, J. G., & Schulenberg, J. E. (2012). Monitoring the Future national survey results on drug use, 1975–2011: Volume II, College students and adults ages 19 –50. Ann Arbor, MI: Institute for Social Research, The University of Michigan.

Kahler, C. W., Strong, D. R., & Read, J. P. (2005). Toward efficient and comprehensive measurement of the alcohol problems continuum in college students: The brief young adult alcohol consequences question- naire. Alcoholism: Clinical and Experimental Research, 29, 1180 –1189. doi:10.1097/01.ALC.0000171940.95813.A5

Lau-Barraco, C., & Collins, R. L. (2011). Social networks and alcohol use among nonstudent emerging adults: A preliminary study. Addictive Behaviors, 36, 47–54. doi:10.1016/j.addbeh.2010.08.017

LeSage, J. P. (2010). Applied econometrics using Matlab. Longabaugh, R., & Zywiak, W. H. (2002). Project COMBINE: A manual

for the administration of the Important People Instrument. Providence, RI: Brown University.

Mackey, E. R., & La Greca, A. M. (2007). Adolescents’ eating, exercise, and weight control behaviors: Does peer crowd affiliation play a role? Journal of Pediatric Psychology, 32, 13–23. doi:10.1093/jpepsy/jsl041

McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social networks. Annual Review of Sociology, 27, 415– 444. doi:10.1146/annurev.soc.27.1.415

Miller, R. G. (1981). Simultaneous Statistical Inference. New York, NY: Springer. doi:10.1007/978-1-4613-8122-8

Newman, M., & Girvan, M. (2004). Finding and evaluating community structure in networks. Physical Review E, 69, 026113. doi:10.1103/ PhysRevE.69.026113

Oetting, E. R., & Donnermeyer, J. F. (1998). Primary socialization theory: The etiology of drug use and deviance. I. Substance Use & Misuse, 33, 995–1026. doi:10.3109/10826089809056252

Phua, J. (2011). The influence of peer norms and popularity on smoking and drinking behavior among college fraternity members: A social network analysis. Social Influence, 6, 153–168. doi:10.1080/15534510 .2011.584445

R Development Core Team. (2011). R: A language and environment for statistical computing: R Foundation for Statistical Computing, Vienna, Austria.

Reboussin, B. A., Song, E. Y., & Wolfson, M. (2012). Social influences on the clustering of underage risky drinking and its consequences in com- munities. Journal of Studies on Alcohol and Drugs, 73, 890 – 898.

Reifman, A., Watson, W., & McCourt, A. (2006). Social networks and college drinking: Probing processes of social influence and selection. Personality and Social Psychology Bulletin, 32, 820 – 832. doi:10.1177/ 0146167206286219

Rivis, A., & Sheeran, P. (2003). Descriptive norms as an additional predictor in the Theory of Planned Behaviour: A meta-analysis. Current Psychology: A Journal for Diverse Perspectives on Diverse Psycholog- ical Issues, 22, 218 –233.

Sallis, J. F., Taylor, W. C., Dowda, M., Freedson, P. S., & Pate, R. R. (2002). Correlates of vigorous physical activity for children in grades 1 through 12: Comparing parent-reported and objectively measured phys- ical activity. Pediatric Exercise Science, 14, 30 – 44.

Steglich, C. E. G., Snijders, T. A. B., & Pearson, M. (2010). Dynamic networks and behavior: Separating selection from influence. Sociologi- cal Methodology, 40, 329 –393. doi:10.1111/j.1467-9531.2010.01225.x

Terry-McElrath, Y. M., & O’Malley, P. M. (2011). Substance use and exercise participation among young adults: Parallel trajectories in a national cohort-sequential study. Addiction, 106, 1855–1865. doi: 10.1111/j.1360-0443.2011.03489.x

U.S. Department of Health and Human Services. (2008). U.S. physical activity statistics. Retrieved from http://apps.nccd.cdc.gov/PASurveillance/ DemoCompareResultV.asp?Year�2007&State�1&Cat�1&CI�#result

Valente, T. W., Ritt-Olson, A., Stacy, A., Unger, J. B., Okamoto, J., & Sussman, S. (2007). Peer acceleration: Effects of a social network tailored substance abuse prevention program among high-risk adoles- cents. Addiction, 102, 1804 –1815. doi:10.1111/j.1360-0443.2007 .01992.x

Voorhees, C. C., Murray, D., Welk, G., Birnbaum, A., Ribisl, K. M., Johnson, C. C., . . . Jobe, J. B. (2005). The role of peer social network factors and physical activity in adolescent girls. American Journal of Health Behavior, 29, 183–190. doi:10.5993/AJHB.29.2.9

Wasserman, S., & Faust, K. (1994). Social Network Analysis: Methods and Applications. Cambridge, UK: Cambridge University Press. doi: 10.1017/CBO9780511815478

Wechsler, H., Dowdall, G. W., Davenport, A., & Castillo, S. (1995). Correlates of college student binge drinking. American Journal of Public Health, 85, 921–926. doi:10.2105/AJPH.85.7.921

Wood, M. D., Read, J. P., Mitchell, R. E., & Brand, N. H. (2004). Do parents still matter? Parent and peer influences on alcohol involvement among recent high school graduates. Psychology of Addictive Behaviors, 18, 19 –30. doi:10.1037/0893-164X.18.1.19

Received January 30, 2012 Revision received July 26, 2013

Accepted July 30, 2013 �

T hi

s do

cu m

en t

is co

py ri

gh te

d by

th e

A m

er ic

an P

sy ch

ol og

ic al

A ss

oc ia

ti on

or on

e of

it s

al li

ed pu

bl is

he rs

. T

hi s

ar ti

cl e

is in

te nd

ed so

le ly

fo r

th e

pe rs

on al

us e

of th

e in

di vi

du al

us er

an d

is no

t to

be di

ss em

in at

ed br

oa dl

y.

1142 BARNETT ET AL.