Effects and Detection of Selective Modifications in Samples: Multivariate Models and Techniques.
Abstract
The study of tests of censoring of extreme sample values has been the major research topic. In contrast to earlier work, aimed at evaluating the effect of using specific inaccurate estimates of population distributions, attention has been concentrated on finding how little information suffices to make tests feasible. In particular, it has been found that one complete sample, of size comparable to that of a suspected incomplete sample, and known to be from the same population, makes possible a fairly powerful test. Even when it is only known that some samples, out of two or more, are complete, it is possible to construct useful tests. The work was also concerned with extreme values, and the use of order statistics in estimating ratios - a technique which can also be applied to censored samples. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 03, 1977
- Accession Number
- ADA045901
Entities
People
- Norman L. Johnson
Organizations
- University of North Carolina at Chapel Hill