ROBUST ESTIMATES OF ORDERED PARAMETERS WITH RIGHT CENSORED OBSERVATIONS

Let θi denote the mean for population i, i=1,2,,.,k. We would like to estimate each of the means when it is believed a priori that θ1<θ2<…θk. The observations taken from each population are subject to arbitrary right censorhsip. A comparison is made between the Kaplan-Meier median and the trimmed means (α=2,20,30, and 40) when the pool adjacent violators algorithm is applied. The comparison is based on the estimated mean square errors. Various underlying population distributions are considered. © 1990 Taylor & Francis Group, LLC. All rights reserved.