We controlled for adult smoking to rule this out as an alternative explanation for youth smoking (Tyas & Pederson, 1998). Friend Smoking Adolescents were asked to think of their five best friends at school. Next, they indicated for whether any of their friends had ever tried cigarettes. This variable was dummy coded as yes = 1 and no = 0. Friend smoking is the most consistent predictor of youth smoking. Thus, we controlled for friend smoking to rule this out as an alternative explanation for youth smoking (Tyas & Pederson, 1998). Demographic Characteristics Age, gender, and socioeconomic status (SES) were self-reported. We used father��s and mother��s education as indicators of SES.
Students responded to one question (once for their father and once for their mother): ��What is the highest grade completed by your father/mother?�� Response options were 1 = 8th grade or less, 2 = Some high school, 3 = High school graduate, 4 = Some college, 5 = College graduate, and 6 = Advanced degree. Data from students who did not know their parents�� educational level were treated as missing. Analytic Plan We conducted all descriptive analyses with SPSS 19.0. We tested for gender differences with t tests for continuous and ��2 tests for categorical variables. To perform structural equation modeling with latent variables, we used Mplus Version 6.1 (Muth��n & Muth��n, 2010). Missing data were handled in Mplus 6.1 with weighted least squares estimation. Weighted least squares estimation uses all available data, except for missing values on covariates (i.e., age, SES, friend and adult smoking in the current study).
Weighted least squares estimation is superior to other missing data techniques (e.g., list-wise and pair-wise deletion) in terms of model estimation, bias, and efficiency. It is also relatively equivalent to multiple imputation techniques (Asparouhov & Muth��n, 2010). RESULTS Descriptive Analyses Table 1 shows descriptive characteristics for the overall sample (N = 1,436) for girls and boys. The mean age was 13.97 years (SD = 0.4), and boys were slightly older (M = 14.0, SD = 0.4) than girls (M = 13.95, SD = 0.4) (p < .05). Compared with boys, girls had higher mean scores on acculturation (p < .001), enculturation (p < .001), familismo (p < .05), respeto (p < .05), and fatalismo (p < .05). Compared with girls, boys had higher mean scores on everyday discrimination (p < .
05) and traditional gender roles (p < .001). Boys were also more likely to have smoked cigarettes in the past 30 days (p < .001). Table 1. Descriptive Characteristics for Overall Sample, Girls, and Boys Table 2 shows correlations among all constructs. Although many of these correlations were statistically Dacomitinib significant, their magnitude was small to moderate, suggesting low multicollinearity. We also conducted a multicollinearity diagnostic test in version 19 of SPSS which further indicated that multicollinearity was not a problem.