Stochastic covariates in binary regression

Abstract

Binary regression has many medical applications. In applying the technique, the tradition is to assume the risk factor X as a non-stochastic variable. In most situations, however, X is stochastic. In this study, we discuss the case when X is stochastic in nature, which is more realistic from a practical point of view than X being non-stochastic. We show that our solutions are much more precise than those obtained by treating X as non-stochastic when, in fact, it is stochastic.