Back To Index Previous Article Next Article Full Text


Statistica Sinica 7(1997), 771-787


LOG-CONCAVITY AND INEQUALITIES FOR

CHI-SQUARE, F AND BETA DISTRIBUTIONS

WITH APPLICATIONS IN MULTIPLE COMPARISONS


H. Finner and M. Roters


Universität Trier


Abstract: In several recent papers log-concavity results and related inequalities for a variety of distributions were obtained. This work is supposed to derive a nearly complete list of corresponding properties concerning the cdf's and some related functions for Beta as well as for central and non-central Chi-square and F distributions, where hitherto only partial results were available. To this end we introduce a generalized reproductive property, thereby extending the relationships between total positivity of order 2, log-concavity and reproductivity developed in Das Gupta and Sarkar (1984). The key to our results are log-concavity properties of the non-central Chi-square distribution with zero degrees of freedom introduced by Siegel (1979). Finally one of the results for the central F distribution is used to solve a monotonicity problem for a stepwise multiple F-test procedure for all pairwise comparisons of k means.



Key words and phrases: Beta distribution, Chi-square distribution, convolution theorem, doubly non-central F distribution, eccentric part of Chi-square, F distribution, log-concave, log-convex, multiple comparisous, pairwise comparisons, Pólya frequency function, Prekopa's theorem, probability inequality, reproductive property, stepwise multiple F-test, total positivity of order 2.


Back To Index Previous Article Next Article Full Text