重複觀測的類別反應資料時常出現在生物統計的應用領域。在很多的應用中，對重複反應的完整多變量建構模型並不是那麼有興趣，而比較反應變數的第一階的邊際分布，通常是最主要的興趣。本計畫將延伸Klingenberg與Agregti（2006）所提出的重複觀測多變量（因此多變量之間可能是相關的或相依的）且每一單變量為二元反應之結果到多元反應。本計畫將推導華德與分數檢定並探討由Pesarin（2005）提出無母數組合函數（nonparametric combination function）於多變量排列檢定之性質。另對稀疏資料，我們執行模擬分析來比較多變量排列檢定與重抽法近似正確分布之情況。 ; Repeated categorical response data occur commonly in biostatistics applications. In many applications, modeling the full multivariate dependence among repeated response is of less interest than comparing first-order marginal distributions of the response. In this project we will extend the results of Klingenberg and Agresti (2006), which is two repeated measures on multivariate variables with each variable having binary responses. The Wald and score tests will be derived. The properties of nonparametric combination function on multivariate permutation tests proposed by Pesarin (2005) will also be studies. For sparse data, we will conduct a simulation study for approximating the distribution of multivariate permutation tests and the bootstrap method to the exact distribution. ; 研究期間 9708 ~ 9807
