Minimum (h,φ)-divergences estimators with weights

摘要

We consider experiments involving the observation of a discrete random variable, or quantitative classification process and we also assume that in addition to probability of each value or class we know its “utility” or “weight” (or more precisely, we can quantify the “nature” of each value or class). In this paper a procedure of minimum divergence estimation based on (h,φ)-divergences is analyzed, for the considered experiments, and its asymptotic behaviour is studied.