adding distal outcomes to the growth model. The results obtained when restricting the analyses to those with alcohol data were compared with a second bias-adjusted set derived using inverse probability weighting (IPW) [28]. This method was employed in preference to extending the likelihood-based method to cover missingness in the outcome since it was felt less likely that the MAR assumption would hold for these data. For the bias-adjustment, logistic regression models were estimated to predict inclusion in the final sample using a number of indicators of family adversity in childhood (good predictors of subsequent dropout). The reciprocal of the predicted probabilities from these models were then used as sampling weights (ranging from 1.8 to 18.2 for some individuals) to adjust the regression models of interest.
