Hierarchical Regression Analyses.

Note. Sets of predictors were added in sequence from M0 to M7. M0  =  Null Model (intercept only), M1 =  demographic variables (age, gender, highest education achieved, employment status, Maasai vs. other ethnicity), M2 =  village membership, M3 =  weekly media consumption (radio, television, print media), M4 =  HIV knowledge, M5 =  psychographic factors. Change in model fit was assessed via likelihood ratio testing: i.e., change in deviance relative to change in degrees of freedom. Sets shown to improve model fit (* p<.05) were carried forward in subsequent analyses.

a

Deviance (-2*log-likelihood) is reported for each model. R² is reported in brackets for best-fitting logistic models.

b

Analyses were carried out on data from the subset of individuals who reported previous sexual activity.

c

Age was not included in the set of basic demographic variables in this analysis.