Title

Critical values for multiple testing and comparisons: one step and step down procedures

Authors

Authors

P. N. Somerville

Comments

Authors: contact us about adding a copy of your work at STARS@ucf.edu

Abbreviated Journal Title

J. Stat. Plan. Infer.

Keywords

multiple comparisons; multiple testing; fortran programs; correlated; data; critical values; SIMULTANEOUS CONFIDENCE-INTERVALS; Statistics & Probability

Abstract

The methodology developed in Somerville (Proceedings of the 25th Symposium on the Interface, Computing Science and Statistics, April 1993, pp. 352-356; Technical Report TR-94-1, Department of Statistics, University of Central Florida, 1994; Comput. Statist. Data Anal. 25, 1997. 217-233; J. Comput. Graph. Stat. 7 (4), 1998, 529-544) to calculate the constants necessary for one-step and step-down multiple testing, and its computer implementations are reviewed in this paper. The constants can be calculated for arbitrary variance-covariance matrices, arbitrary numbers of populations, and arbitrary number of degrees of freedom for sigma(2). The methodology is based on methods for calculating values for multivariate normal and multivariate-t integrals over certain convex regions. A comparison of the procedures of Edwards and Berry (Biometrics 43, 1987, 913-928) and Somerville (1997) is given with the latter procedure indicated to be orders of magnitude faster for most practical situations. In addition, an example is given, demonstrating that ignoring correlations of parameter estimates can result in multiple tests for comparisons which are very conservative. (C) 1999 Elsevier Science B.V. All rights reserved. MSC: 62J15; 65C05; 65D20; 65D30.

Journal Title

Journal of Statistical Planning and Inference

Volume

82

Issue/Number

1-2

Publication Date

1-1-1999

Document Type

Article

Language

English

First Page

129

Last Page

138

WOS Identifier

WOS:000083580500010

ISSN

0378-3758

Share

COinS