調節複迴歸的運用

Abstract

本研究一開始在說明調節複迴歸的型式和運用方向，並且說明了可以運用在調節複迴歸的檢定方式，接著將就所提出來的方法（Welch T-test、WLS、MWLS）來提出一些討論並詳細說明能運用的條件以及分布情形。 本文中最主要的就是Welch T-test和WLS之間公式的推導，利用公式的推導來證明兩個檢定式的長像是相同的，但是因為兩個的自由度不同，所以檢定值的分佈也會有所不同。 最後在利用SAS軟體來進行模擬分析的動作，分別利用樣本數的大小和數目的異同，來分別做了4組的實驗，而四組當中又將不同的變異數做各種的組合，求出在各種情形下的3種檢定的型一誤差，並藉此比較出在α＝0.05的情形下，Welch T-test優於其他2種檢定方法的證明，並且利用公式推倒的證明和模擬分析的研究結果來推翻了Overton（2001）所提出的說法。

This research is stating the modeling of Moderates Multiple Regression and application direction at the beginning, and has proved that can use it in the examination way of Moderates Multiple Regression , then will put forward the condition which can be used in some discussions and elaboration and distribution situation on the method (Welch T-test , WLS , MWLS ) put forward. A main one in this research is Welch T-test and WLS formula derive in the article originally, is it prove two assay type long as if the same to make use of the deriving of the formula, but because the degree of freedom of two is different, the distribution of examining definite value will be different to some extent. Finally using the SAS software to imitate movements analyzed, utilize the similarities and differences between size and figure that samples are being counted separately, we use 4 group of experiments separately, and four group count various kinds of associations different variation, type one error that the eggplant lies in the types of 3 kinds of examination under various kinds of situations, and relatively lie in α= 0.05 situations, Welch T-test is superior to the identification of 2 kinds of other examination methods , and the result of study of utilizing identification and simulation that formulae pushed over to analyses has overthrown the statements put forward of Overton (2001 ).