Summary: | 碩士 === 輔仁大學 === 統計資訊學系應用統計碩士在職專班 === 104 === The methods for combining tests have been widely used in various meta-analyses. One of the commonly used methods is the Fisher combination method, which combines the p-values of different tests in a logarithmic transformed manner. Basically, there are two types of combination methods, the quantile combination methods and truncated combination methods. Among the quantile combination methods, many studies have shown that the inverse chi-square method has better power performance than inverse normal method. On the other hand, among the truncated combination methods, it is demonstrated that the rank truncated product method is superior to Fisher’s method and truncated product method. In the light of these results, this study aims to combine the rank truncated product method and the inverse chi-square method to establish a new method. In this thesis, experimental simulations are performed to evaluate the power performance of the proposed method.
|