Trimmed weighted Simes' test for two one-sided hypotheses with arbitrarily correlated test statistics.

摘要: The two-sided Simes test is known to control the type I error rate with bivariate normal statistics. For one-sided hypotheses, of requires that correlation between statistics non-negative. In this article, we introduce a trimmed version weighted for two hypotheses which rejects if (i) and (ii) both p-values are below one minus respective Bonferroni adjusted level. We show controls at nominal significance level α common distribution point symmetric 2α These assumptions apply, instance, arbitrary correlation. simulation study, compare power untrimmed test. An additional result article ensures usual under weak positive regression dependence condition case hypotheses. This shown apply one- or two-sample t-tests endpoints corresponding such types has not been considered before. According our main result, then also applies t-test