For the second post‐hoc analysis, logit models were constructed to estimate the probability of transition to non‐drinking during the life‐course within each baseline category of current drinkers. Sex‐specific binary variables were coded for each such category according to whether or not constituent participants had transitioned to non‐drinking at a given phase of observation. The ‐xtlogit‐ command was used to predict the probability of transition to non‐drinking within each baseline consumption category, with the predicted probabilities then plotted as a function of age 23.
