Parameter Shrinkage for Age-Period-Cohort Modeling of Opioid Mortality Rates

Statistical methods that shrink parameters towards zero can produce lower predictive variance than does maximum likelihood. This paper discusses an approach to doing this for age-period-cohort models, and applies it to fitting opioid mortality rates with a generalization of the Lee-Carter model including cohorts. Bayesian parameter shrinkage has some practical advantages over classical versions