Fisher information in censored samples from folded and unfolded populations

Abstract

Fisher information (FI) forms the backbone for many parametric inferential procedures and provides a useful metric for the design of experiments. The purpose of this paper is to suggest an easy way to compute the FI in censored samples from an unfolded symmetric distribution and its folded version with minimal computation that involves only the expectations of functions of order statistics from the folded distribution. In particular we obtain expressions for the FI in a single order statistic and in Type-II censored samples from an unfolded distribution and the associated folded distribution. We illustrate our results by computing the FI on the scale parameter in censored samples from a Laplace (double exponential) distribution in terms of the expectations of special functions of order statistics from exponential samples. We discuss the limiting forms and illustrate applications of our results.