Short notes on maximum likelihood inference for control-based pattern-mixture models

Abstract

A likelihood-based analytical approach has been proposed for the control-based pattern-mixture model and its extension. In this note, we derive equivalent but simpler analytical expressions for the treatment effect and its variance for these control-based pattern mixture models. Our formulae are easier to use and interpret. An application of our formulae to an antidepressant trial is provided, in which the likelihood-based analysis is compared with the multiple imputation approach.