On Combining Independent Tests In Case Of Conditional Logistic Distribution

Abstract

Combining $n$ independent tests of simple hypotheses, vs one-tailed alternative as $n$ approaches infinity, in case of conditional normal distribution with probability density function $X|\theta \sim \mathscr{L} (\gamma \theta,1)$, $\theta \in [a,\infty),a \geq 0$ for the case where $\theta_1, \theta_2,...$ are distributed according to the distribution function (DF) $G_\theta$ was studied. Four nonparametric combination procedures (Fisher, logistic, sum of P-values and inverse normal) were compared via the exact Bahadur slope.