Model validation

We used repeated data from NESARC-III to externally validate the calibrated transition probabilities. NESARC-III represents an independent source of data because it was not used to develop the transition probabilities.⁶² We used the three sets of transition probabilities: TM₁ for the first 3 years elapsed from the earlier drinking period, TM₂ for the next 5 years, and TM₃ for remaining years up to the survey period. We stratified the validation dataset and visually examined predicted versus actual state distributions by gender, age, race/ethnicity, initial drinking state, and length of time between the earlier drinking period and the survey period.