
Title
Fisher Information In Incomplete Univariate And Bivariate Samples
Speaker
H. N. Nagaraja, The Ohio State University
Abstract
We consider three types of incomplete data generated from a random sample of size n from either a univariate population X or a bivariate population (X,Y). They are: (a) A collection of X order statistics, in particular, Type II censored samples, (b) All X upper (or lower) records and associated record times when the n observations are presented sequentially, and (c) A collection of X order statistics and their associated Y concomitants. We describe interesting properties of the Fisher Information regarding a parameter (scalar or vector) for such data sets. We discuss computational issues, and applications of our results to inference on the parameter. Our examples include exponential, (univariate and bivariate) normal, and Farlie-Gumbel-Morgenstern distributions.