During the Tofacitinib JAK3 spring of 12th grade, the average age was 18.15 years (SD=0.34). Some 30% of participants received free or reduced-price lunch in the first 2 years of the study. Measures All the variables were based on self-reports from the youth. Self-reports of smoking have been shown to be valid in most studies (Patrick et al., 1994). Participants were asked to report on the number of cigarettes smoked per day in the last month: 0 (coded 1), less than 1 (coded 2), 1�C5 (coded 3), about a half-pack (coded 4), a pack (coded 5), or more than a pack (coded 6). Participants were divided into nonsmokers (coded 1), light and intermittent smokers (coded 2 and 3), and heavy smokers (coded 4 or higher). Age at onset was ascertained from annual prevalence data and was the first year that a participant reported smoking.

Gender was coded 1 for males (n=525) and 0 for females (n=465). College status was assessed at F1 and coded 1 for participants enrolled part time or full time in a 2-year or 4-year college (n=381) and 0 for those not enrolled in college (or still in high school; n=488). Most (81%) of the college students were full-time students, and only a small percentage of those classified as college students were part-time students who also held full-time jobs (less than 10% of all college students). We measured binge drinking as a time-varying covariate. Frequency of binge drinking was the number of times males drank five or more and females drank four or more alcoholic drinks in a row in the prior 30 days (Wechsler, Lee, Kuo, & Lee, 2000).

Participants were trichotomized as non�Cbinge drinkers (zero times), infrequent binge drinkers (one to two times), and frequent binge drinkers (more than three times; Wechsler et al., 2000). Data analyses We used Markov models to examine within-individual change in smoking stage membership (nonsmoking, light and intermittent smoking, and heavy smoking) over time. Markov models (Collins, Graham, Long, & Hansen, 1994; Collins & Wugalter, 1992) have become widely used in substance use research when substance use is conceptualized as discrete stages in a developmental process (Jackson, Sher, Gotham, & Wood, 2001). Although such models have commonly taken the form of hidden Markov models (sometimes called ��latent transition�� models), where stages are measured with multiple indicators, we used a Markov model in which stages are based on a single measure of smoking (quantity per month) with the assumption of no measurement error.

These models have advantages over other person-centered techniques when there is a high degree of movement into and out of behavior states. Survival analysis is an alternative person-centered technique Drug_discovery that could be used to quantify, describe, and predict transitions into light and intermittent smoking, but it does not provide a way to simultaneously model movement into and out of light and intermittent smoking over time.