Small Sample Comparisons of Exponential Reliability Estimates for Type I Censoring.
Abstract
Our study has shown that using X squared sub 2 c + 2 to estimate theta and R for small samples sizes, produces estimates that are too conservative when compared to Bartholomew's exact method. If X squared sub 2 c is used to estimate theta and R, optimistic estimates will be obtained. Better estimates can be obtained for small samples sizes by using X squared sub 2 c + 1. Therefore, for small sample sizes, we recommended using X squared sub 2 c + 1 instead of X squared sub 2 c + 2 or X squared sub 2 c. During our study, much to our surprise, we found that the lower confidence limits could not be computed for all realistic values of theta in Bartholomew's exact method. This occurred when theta approached its maximum theoretical value for very small sample sizes. An explanation for this has not yet been determined by the authors.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1987
- Accession Number
- ADA190021
Entities
People
- Jerry Thomas
- Linda L. Crawford
Organizations
- Ballistic Research Laboratory