Accession Number : AD0724806
Title : On a Family of Lesser Known Goodness of Fit Criteria.
Descriptive Note : Themis optimization research program,
Corporate Author : TEXAS A AND M UNIV COLLEGE STATION INST OF STATISTICS
Personal Author(s) : Hartley,H. O. ; Pfaffenberger,R. C.
Report Date : MAY 1971
Pagination or Media Count : 26
Abstract : The recently developed statistical theory for setting confidence limits for the extrema in mathematical programming has necessitated certain extensions of the statistical methodology available for this purpose. The statistical problem is that of setting confidence limits for the end point of a finite range distribution given a sample of n values drawn from it. Such confidence limits require the use of the so-called 'non-parametric' 'goodness of fit' criteria for the comparison of sample and c.d.f. The criteria available in the literature were all found to be unsatisfactory with regard to detecting departures in the 'tails of the distribution'. It was therefore necessary to develop a considerably more powerful family of criteria in order to obtain satisfactory confidence points. The present report provides the derivation and the complete theory for these new criteria. (Author)
Descriptors : (*STATISTICAL TESTS, SAMPLING), (*MATHEMATICAL PROGRAMMING, CONFIDENCE LIMITS), STATISTICAL DISTRIBUTIONS, MONTE CARLO METHOD, APPROXIMATION(MATHEMATICS), MULTIVARIATE ANALYSIS
Subject Categories : Statistics and Probability
Distribution Statement : APPROVED FOR PUBLIC RELEASE