Table 2 documents estimates of the partialled predictive power of the demographic and high school risk factors for binge drinking and extreme binge drinking, based on multivariate logistic regression models. Missing data dummy variables were included for all predictors with missing data, with means also assigned for continuous predictors. Students with missing data on gender were more likely to binge drink at all levels. Students with missing data on high school grades were less likely to engage in 5+ binge. Those with missing data on college plans were more likely to engage in 5+ binge. Students with missing data on days cut school were more likely to report 15+ binge. Students with missing data on disapproval of binge drinking were more likely to engage in all levels of binge drinking. Those with missing data on perceived risk were less likely to report 5+ binge drinking. Students with missing data on cigarette or marijuana use had greater odds of 15+ binge drinking. There were no significant differences for students with missing data on race/ethnicity, parental education, religiosity, evenings out, or friends get drunk compared to students who gave valid data for these measures.
