Publication Date

5-14-1969

Abstract

Necessary and sufficient conditions are given for reversing Jensen's inequality for continuous convex functions, and the result is then stated for the discrete case. The main theorem of Hanson and Koopmans' (1964) pioneering tolerance limit paper is extended to the case of any finite number of order statistics from a probability distribution F for which either -log(1 - F) or - log F is convex. In the first case upper tolerance limits and in the second case lower tolerance limits are exhibited for any content p and any confidence ɣ, 0 < p, ɣ < 1 whenever the sample size is not less than two. When F is such that both functions are convex, two-sided, p-content tolerance intervals at confidence level rare constructed for any sample size greater than one. The distributions for which both functions are convex include the Pólya distribution functions of order two, which includes the normal and exponential families. The distribution of the coverage of the tolerance limits is calculated using an extension of Renyi’s (1953) work on exponential random variables. Goodman and Madansky's (1962) most stable criterion for tolerance intervals is characterized in terms of its second moment about a point, and some of its other properties are established. The results are applied to develop a method for finding the most stable one-sided tolerance limits. The general method is applicable to the problem of finding the most stable two-sided tolerance intervals, but the required explicit calculation of the probability distribution function is lacking. Instead the two-sided tolerance interval is found in terms of two one-sided tolerance limits.

Degree Name

Mathematics

Level of Degree

Doctoral

Department Name

Mathematics & Statistics

First Committee Member (Chair)

Lambert Herman Koopmans

Second Committee Member

Julius Rubin Blum

Third Committee Member

Clifford Ray Qualls

Fourth Committee Member

Liang-Shin Hahn

Language

English

Document Type

Dissertation

Download

Share

COinS