All covariates that had a
statistically significant bivariate association with exposure to movie smoking were included in the multivariate
model, for which the β coefficients were
equivalent to the number of additional episodes of movie smoking attributable to that factor.