Information Theory and Maximum Product of Spacings Estimation

Abstract

A principle of least information is proposed for estimating continuous univariate distributions. Data are represented by quantiles; the principle yields a reference distribution that is identical with the maximum product of spacings estimate.