ESTIMATION UNDER INEQUALITY CONSTRAINTS

Abstract

     

In this article we make à comparative study of the classical and Bayesian approaches to the problem of estimation under inequality constraints. The problem arises in several contexts. A statistician trying to fit a probability model to a given set of observations may have prior information, based upon personal belief Ôr past studies, that the parameters of interest lie in a certain subset of the parameter space. If he chooses to utilize this information, he has a problem of estimation under inequality constraints sometimes, the constraints are implied by theoretical assumptions underlying the assumed model. In nested design models, for example, the components of the error terms are assumed to be independent normal variates, which imply that the observations generated by the model are positively correlated.