Combining Independent Statistical Tests
Margareta Westberg · Journal of the Royal Statistical Society Series D (The Statistician) · 1985
In the present study two well-known combination methods, Fisher's and Tippett's, are compared according to their power. The calculations are made for normally and chi-square distributed test statistics. None of the two procedures is uniformly better than the other according to the power but sometimes the power curves cross each other. The calculated power-graphs give guidelines for when to use Fisher's method and when to use Tippett's.