| In a case control study, the Armitage's trend test is more powerful for detecting association of association of a candidate gene with a disease than the chi-square 2X3 test of equal proportions. But, no studies have been reported on the effect of classification errors on the efficiency of the two tests.;We derive the non-centrality parameter for Armitage's test in the presence of errors, calculate which classification errors are most costly in terms of increased sample size to maintain constant power and significance level for case-control studies of genetic association, and finally compare the power of Armitage's trend test with one of chi-square homogeneity test under various genetic models. We compare three weightings of the trend test to the 2X3 Chi-square test. In the cases examined, there is always a weighted trend test that has greatest power. The weighting, however, varies with the situation studied, and the 2X3 Chi-square test is always has power close to (but less than) the best trend test. |
