Combining Correlated P-values From Primary Data Analyses

Research results on the same subject, extracted from scientific papers or clinical trials, are combined to determine a consensus. We primarily concerned with combining p-values experiments that may be correlated. have two methods, non-Bayesian method and Bayesian method. use model combine these assume follow certain distribution, for example, chi-square normal. The distribution requires independent identically distributed (iid) random variables. When data correlated non-iid, we cannot such distribution. In order do so, need adjusted, adjustment is done “indirectly” through test statistics. Specifically, one statistic (TS** ) obtained non-iid other (TS) iid data. ratio between statistics adjust (TS**) its violation. adjusted TS** named as “effective statistics” (ETS), which then used statistical inferences assumed As it difficult estimate correlation, provide more coherent p-values, also introduce novel both examples illustrate additional given