Prior to adopting the multinomial logistic regression model (i.e., generalized logistic regression model); each logistic regression model with the 4-level offspring smoking outcome was tested to determine if the assumption of proportional odds was met (i.e., ordinal pattern of the 4-level smoking outcome variable). The test yielded a significant violation of the proportional odds model (p<0.05), so the multinomial logistic regression model was adopted for all analyses.
