Prediction of treatments effects in a biased allocation model

Abstract

Robbins and Zhang [15] provide consistent estimators of multiplicative treatment effects under a biased treatment allocation scheme, and illustrate their methodology within Poisson and binomial models. Here we use predictive criteria to assess the differential treatment effects, and develop predictive distributions for the Poisson errors in variables models. With a hierarchical prior structure, various approximations are investigated, and an illustrative example is included.