Multi‐Stage Estimation Procedures for the Mean OE A U‐Statistic and Some Associated Moderate Sample Size Comparisons

Abstract

ABSTRACT :

In this paper we estimate the mean of a U‐statistic when the distribution in the population from where the independent samples are taken is unspecified. We develop two‐stage and three‐stage minimum risk, with squared error loss plus linear cost, point estimation procedures which are both asymptotically risk efficient. These multistage procedures are also operationally much simpler than Sen and Ghosh's (1981)purely sequential scheme. The performances of these three procedures are then compared via computer simulation for mode rate sample sizes under a few specific types of population distributions.

1980 AMS Subject classifications: 62Ll2, 62G05.