Advances and Applications in Statistics
Volume 5, Issue 1, Pages 91 - 110
(April 2005)
|
|
PERMUTATION TESTS IN SCREENING TWO-LEVEL FACTORIAL EXPERIMENTS
Luigi Salmaso (Italy)
|
Abstract: In the literature there are some recent proposals for testing for active effects in factorial designs. A comprehensive review is provided by Hamada and Balakrishnan [Statist. Sinica 8 (1998), 1-41], in the case of unreplicated factorials. In this paper, we first review the two permutation testing procedures from the literature, also providing an extension of the paired permutation procedure and a comparative simulation study when the experimental design is a full factorial. Then, we introduce a new "permutation" approach for testing for active effects in Hadamard matrices. The idea underlying the permutation process is innovative in the sense that rather than permuting responses of a single observed design matrix (or, equivalently, permuting rows of the design matrix keeping the responses fixed), we exchange inequivalent matrices in order to obtain the permutation distribution. This test, called an IMPT (Inequivalent Matrices Permutation Test), behaves well both under the null hypothesis and in power, even with heavy-tailed distributions. |
Keywords and phrases: inequivalent matrices, Hadamard matrices, paired permutations, realigning observations, restricted randomization. |
|
Number of Downloads: 363 | Number of Views: 1377 |
|