Ranking Procedures for Arbitrarily Restricted Observation

Publisher Website
Google Scholar
Abstract
It is shown that a statistic W proposed by Gehan [1965a] for comparing 2 sets of arbitrarily censored data, can be represented as the total of a sample from a readily defined finite population of scores: this approach simplifies both the method of a computation and the determination of the permutation distribution of W and its variance, as well as justifying an assumption of asymptotic normality. A possible modification of the test by applying the Wilcoxon rank-sum procedure to the sample of scores is discussed. Possible patterns of censoring are considered, and criteria are given for deciding whether any 2 samples can be compared without bias.

This publication has 0 references indexed in Scilit: