A Goodness of Fit Test Based on the Spacing of Selected Order Statistics.
Abstract
Work on the statistical control of the optima of mathematical functions (statistical solution of non-linear optimization) has lead to the statistical problem of estimating the upper limit of a finite range distribution from a random sample of ascent paths. The task of setting confidence limits for this upper limit necessitates the use of a non-parametric goodness of fit criterion. Since the criteria in the literature are not particularly suited to this purpose, it has been necessary to develop suitable criteria ab initio. Specifically, the present report is concerned with the development of a one sample goodness of fit test which is particularly sensitive to tail discrepancies between the true distribution F(x) and the hypothesized distribution F sub 0(x) of a continuous variate x. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1971
- Accession Number
- AD0733052
Entities
People
- Dan Lurie
- Herman Otto Hartley
Organizations
- Texas A&M University