Abstract
Analysis of the K ordered odds ratio parameters, namely the analysis of 3-way interaction in the 2×2×K contingency table is a very common problem in the epidemiology and also in the clinical trial. Recently Hirotsu et al.(2001) proposed the max accumulatedx2 statistic for a 2×J×K table related with the association analysis of disease and bivariate allele frequencies, which includes the 2×2×K table as its important special case. The purpose of the present paper is to reformulate the max accumulated statistic as useful specifically for the 2×2×K table and compare its power with the restricted likelihood ratio test proposed by Barmi(1997). It is shown that the behavior is rather similar but the max accumulated x2 is easier to handle giving an exact algorithm for the probability calculation for the moderate sample size.
