We reported weighted percentages and directly standardized
rates of alcohol use and other outcomes of interest for
all analyses. The details of the weighting and standardization
procedure are described elsewhere.2 We used chi-square
tests and logistic regression to examine comparisons of rates
of student characteristics and outcomes of interest. We used
the multiple logistic regression technique to assess the relationship
among the outcomes of interest, adjusting for other
covariates. Odds ratios (ORs) and 95% confidence intervals
(CIs) are reported. We used the generalized estimating equations
(GEE) approach to fitting the logistic regression models
to account appropriately for clustered outcomes arising
in our sampling scheme.24,25 All analyses were conducted
using the SAS statistical software packag