Is ‘Personalized’ Optimized? Assessing the Performance of Person-Specific Prediction Models of Alcohol Consumption.

We applied spike-and-slab regression to time series data to assess the ability of personalized models to predict alcohol consumption. In two exemplar cases, the best-performing models predicted alcohol consumption well (AUCs = .81 & .98). However, there were non-significant differences in performance among the top 10 models for each individual.