Analyses began by testing if parent, sibling and peer variables had a bivariate association with the 4 level offspring smoking outcome. All variables significant at the bivariate level were included in the multivariate multinomial logistic regression models. All models controlled for genetic vulnerability for ND imparted from parental smoking. This was accomplished by modeling the effect of the 4-Group design variable described in section 2.2.2. This variable was forced into the model as a dummy variable representing varying degrees of genetic and environmental vulnerability for nicotine dependence. However, we note that the maternal nicotine dependence variable was treated as a covariate and not a measure of genetic and environmental vulnerability because mothers were not twins and thus varying levels of genetic vs. environmental influence could not be established for their influence on offspring smoking behaviors.
