Statistical analyses
Valid questionnaires were entered into a SPSS database. After
that, responses were analyzed for consistency. All estimates
were adjusted using sampling weights to represent the entire
college student population of Brazil. Descriptive and inferential
analyses were conducted using the R (version 2.12.0)
software package. The gender- and age-related effects of
drug use and the ASSIST-WHO scores were assessed using
Wald’s tests. Null hypotheses were rejected when p < 0.05.
All results are expressed as the mean and standard error
(SE). Afterward, logistic regression models were developed to
evaluate whether the effects of gender, age and their interaction
explained the use of each drug for which a distribution
difference was detected by Wald’s test. In each model, the
response variables were either drug use in the last 30 days or
the ASSIST-WHO score. In addition to gender and age, other
sociodemographic, academic and administrative variables
were included as covariates. These other variables included
Brazilian administrative region, the HEI’s administrative organization type, field of study (biological science, humanities
or physical sciences), marital status, socioeconomic status,
ethnicity, and religion affiliation. Variables that did not reach statistical significance were excluded from the models using backward stepwise elimination. Confidence
intervals were estimated using the Bonferroni correction, and
confidence coefficients were set at 95%. The adjusted final
models for each drug are described in detail below.