重複量測資料的層次變換模型

Abstract

變換反應變數的技術在對非服從常態分配的資料常被頻繁地使用。首先，我們回顧三個重要的變換族群例子，他們的值域可能和整個實數線不同。在上述的情況下，我們提出一個針對重複量測資料的變換層次模型。接著，對上述的模型做有母數及半母數的推論。最後，我們也考慮穩健性的推論。

The technique of transforming the response is frequently used for modeling nonnormal data. Three important families of transformations with ranges possibly different from the real line are first reviewed. A transformed hierarchical model is proposed for repeated measurement data in such situations. Both parametric and semiparametric inferences are presented. Consideration of robustness is also discussed.