There may also be complementarity in the usage of marijuana and alcohol

In addition, to the extent that use of one substance affects the usage of another among adolescents, accounting for this interdependence in substance use is important as it can minimize the possibility of obtaining spurious relationships and possibly biased model estimates. While some studies indicate that cigarette smoking is a strong predictor of the concurrent or subsequent usage of alcohol and marijuana, other studies find that alcohol use increases the likelihood of cigarette smoking. In other research, a mutually reinforcing relationship was detected between adolescent alcohol use and smoking, and between cigarette smoking and marijuana use during adolescence and young adulthood. In contrast, other research found that previous use of alcohol did not predict the initiation of marijuana use. Existing studies also indicate that adolescent users of marijuana frequently smoke cigarettes, either as a substitute when marijuana is scarce, or as a means of counteracting the sedating effects of marijuana. The complementary usage of tobacco and marijuana in adolescence may contribute to the eventual dependence on nicotine. The complementary usage of these two substances might be the greatest public health consequence from marijuana use in adolescence.The observed high correlation of alcohol and marijuana use may be due to a shared genetic risk for drug use. On the other hand,motel grow rack this observed correlation may be due to substituting one substance for another as a means of minimizing marijuana withdrawal symptoms.

In one study, daily marijuana users who underwent a period of abstinence drank more often if they had a previous diagnosis of alcohol abuse or dependence. The importance of peers in the transmission of substance use behavior within adolescent friendship networks has given rise to a body of literature which focuses on how social networks can spread substance use behavior. These studies focus on how the co-evolution of adolescent friendship networks and substance use gives rise to peer influence and selection effects within adolescent networks regarding a specific substance use behavior . Peer influence is a type of social influence, and the latter has been theorized from numerous points of view including the Dynamic Social Impact Theory, which states that individuals will become more like those who are socially proximal and as a result their attributes will be correlated. Given the importance of concurrent or sequential usage of cigarettes, alcohol, and marijuana, the purpose of this study is to examine the co-evolution of use of these substances within the dynamic landscape of adolescent friendship networks, which are a primary socialization context for adolescent substance use. A key methodological and theoretical challenge herein is that the context of peer networks must be taken in consideration when studying adolescents’ interdependent substance use, because interpersonal association via peer influence or friendship selection likely shapes the concurrent or sequential use of substances. Not taking into account such peer network effects can result in biased estimates of the interdependence of substance use behaviors, or likewise, the effects of network influence.

Fig 1 displays a hypothetical simple 2-person world in which the “true” model are the solid lines and the dashed lines show possibly spurious intra-personal effects of one substance use on another. This figure is informed by the Dynamic Social Impact Theory, as we posit that person 1 will become more like his or her peer, person 2, because of peer influence via modeling, shared opportunities, social proximity and the consolidation of attitudes and behaviors that may take place in adolescent friendship networks. This social process is captured in pathways from person 1 smoking to person 2 smoking, and in addition from person 1 drinking to person 2 drinking. Note that certain person-specific covariates affect an adolescent’s usage of each substance: this will therefore lead to a correlation in usage across substances for the person. Not accounting for the across-person effects as shown in dotted lines–for example, how the smoking behavior of person 1 affects the smoking behavior of person 2 through a social influence or a selection effect–will result in these correlations being inappropriately captured by the dashed paths. Such correlations would be spurious, thus highlighting why it is critical to account for these network effects when studying concurrent substance use behavior. Although we do not show them here , this figure could also represent pathways linking person 1 smoking and person 2 marijuana use, and analogously, person 1 drinking and person 2 marijuana use. These pathways may be a result of normative processes. The subjective norms construct from the Theory of Planned Behavior, which is the composite of the belief about whether most people approve or disapprove of a behavior and their corresponding motivation to comply with those important referents in their social environment, informs these normative pathways in our model.

Adolescents who are smoking cigarettes, may through such normative beliefs, reinforce the use of other substances among their friends as they display pro-substance use norms and thus their friends may be motivated to comply with their attitudes and behavior, which would indirectly increase friends’ acceptability of using marijuana or alcohol. We are aware of just two existing studies that have simultaneously studied adolescent social networks and the use of more than two substances. One longitudinal study focused upon smoking, drinking, and marijuana use in a sample of 129 Scottish youth finding that while there were statistically significant peer influence effects on alcohol and marijuana use but not on smoking behavior, marijuana users smoked cigarettes more over time. However, the other study, detected neither interdependent association effects nor peer influence effects on cigarette, alcohol, and marijuana use behaviors in a longitudinal study of a sample of US school students. Note that the first study utilized a relatively small sample and the second study is limited to two waves of data, thus diminishing the statistical power of each study. Finally, social influence may not necessarily have symmetric effects on initiation and cessation of substance use. For example, Haas and Schaefer explored this idea in the context of smoking, and found some evidence that influence effects may have a stronger effect on starting smoking behavior, but weaker effects on stopping it. Although we do not have specific hypotheses regarding how such influence effects might operate for other substances such as alcohol or marijuana use, we nonetheless test this asymmetry possibility here in our analyses. Building on past studies focusing on concurrent or sequential substance use in adolescence, we examine the co-evolution of adolescent friendship network ties and whether there was interdependence in usage of cigarettes, alcohol, and marijuana among 3,128 adolescents in two large schools. We utilize three waves of social network data from the National Longitudinal Study of Adolescent to Adult Health. Ecological models of human development informed the conceptualization of this study by situating adolescents in key social contexts exerting primary socialization forces including peer selection, peer influence, and parental influences. This study is also informed by the Dynamic Social Impact Theory, which forms the basis for why youths’ behaviors will be correlated. Lastly, normative constructs from the Theory of Planned Behavior guide our examination of the normative model pathways under study.The data utilized in this study come from early waves of the Add Health study. The respondent record/information was anonymized and de-identified prior to analysis. This study was reviewed and granted approval under exempt review by the Institutional Review Board at the University of California, Irvine. This study does not employ human subjects directly,rolling grow trays as our analyses utilize secondary data, which are de-identified. Written informed consent was given by participants for their answers to be used in this study. We construct separate samples for the two large saturation sample schools, one suburban Northeast public high school referred to as “Sunshine High”, and one rural Midwest public high school referred to as “Jefferson High”. Our data come from the Add Health In-School Survey , the wave 1 In-Home Survey , and the wave 2 In-Home Survey.

Therefore, the average time spans between wave 1 and wave 2 are 8.8 and 7.7 months for students in Sunshine High and Jefferson High, respectively. The average time spans between wave 2 and wave 3 are 10.9 and 11.1 months for students in Sunshine High and Jefferson High, respectively.We utilize the R-based Simulation Investigation of Empirical Network Analysis software package to estimate Stochastic Actor-Based models. We specify each model with three behavior equations in which we focus on how usage of one substance is influenced by the usage of the other two substances, along with one network equation in which we model the network evolution in tie formation and dissolution among adolescents in the school. We estimated the model separately on each school. Besides the key mechanisms illustrated in Fig 1, we adopt a forward selection approach for each parameter via score-type test. In the behavior equations, the linear and quadratic effects capture the time trend of each substance use behavior; peer influence effects are measured as the sum of negative absolute difference between ego’s and alters’ behavior averaged by ego’s out-degree. Additional covariates such as in-degree, parental support, parental monitoring, race , and depressive symptoms are added given that they have been shown to be important covariates in the existing literature, and given that the results from score-type tests reject the null hypothesis that their parameters are 0s. In-degree is important to test, given the debate in the existing literature about the importance of network centrality, or popularity, for explaining substance use. We also control for the effects of how ego’s use of one substance was influenced by alters’ use of two other substances. In the network equation, we include endogenous network effects and homophily selection effects for each substance use behavior as well as additional covariates such as race , gender, grade, and parental education as the results from score-type tests suggest to do so. 501 students in Sunshine High and 166 students in Jefferson High were 12th-graders at t1 and t2and graduated at t3 . These 667 students were constructed as structural zeroes in the networks during the last wave. Due to a survey implementation error in Add Health, some adolescents could only nominate one female and one male friend at t2and t3 . We account for this with a limited nomination variable in the network equation. A Method of Moments estimation is used to estimate the behavior and network parameters in each model so that the target statistics in behaviors and networks can be most accurately calculated. We assess satisfactory model convergence with criteria of t statistics for deviations from targets and the overall maximum convergence ratio . The results of a post hoc time heterogeneity test for the models found no evidence that the co-evolution of substance use behaviors and friendship networks was significantly different across the two time periods, providing no indication of estimation or specification problems. We also perform goodness-of-fit testing for key network statistics in both schools, and display the results in the S1 File. Besides the main SAB model for each school sample, we estimate ancillary models that test whether the interdependent effects are symmetric in increasing and decreasing substance use. This is accomplished by differentiating the “creation” function and the “endowment” function in RSiena. This technique has been applied to explore the asymmetric peer influence effect on adolescent smoking initiation and cessation.A methodological challenge we face is that whereas the questions about smoking and drinking behavior were asked at all three waves, questions about marijuana use were only asked at t2 and t3 . One approach would discard all the information at t1 , but this strategy will reduce the efficiency of analysis, increase standard errors, and decrease statistical power. Instead, we reconstruct adolescent marijuana use at t1 based on four questions. Fig 2 provides a flow chart of the logic, and shows that we in fact have a considerable amount of information that can help us reconstruct probable values for the vast majority of the adolescents. First, if an adolescent has never tried marijuana at t2 , s/he would not have used it at t1, so we can safely code them as a zero at t1 . Next, if an adolescent has tried marijuana at t2 but the age at which he or she tried was above his or her age at t1 , s/he would not have reported using it at t1, so we can safely code them as a zero at t1 . Finally, if an adolescent has tried marijuana at t2 and the age of usage was below his or her age at t1 , we utilize information from two questions “During your life, how many times have you used marijuana?” and “During the past 30 days, how many times did you use marijuana?” at t2.