多重迴歸分析 Multiple Regression Correlation Analysis

Multiple regression correlation analysis is intended for finding the extent to which independent variables have influenced on the dependent variables. The basic MRC can tell you direct effect, indirect effect, spurious effect, total effect, partial correlation coefficients and semipartial correlation coefficients, etc. You might feel it so complicated. To test hypotheses, MRC is a core course in science departments. Advanced MRC involves more independent (IVs) and dependent variables (DVs). For instance, a study is conducted by examining the product effect and user attitude. It is incorrect to analyze the DVs individually, because it will inflate the significance and increase standard error. It is called multicollinearity. For this reason, the MANCOVA (Multivariate Analysis of Covariance) and MANOVA (Multivariate Analysis of Variance) are created to test simultaneously differences among groups of multiple dependent variables. However, according to Cohen, et. al. (2003), with large number of IVs there is the risk of finding something that is not there and not finding something that is there. As a result, less is more means more statistical test validity, more power, and more meaningful results. From the perspective of science, there exist serious flaws that Taiwanese put different elections in a poll. Unfortunately, few of them realize this problem. What are your opinions on this issue?
多重迴歸分析主要是用於分析各種因素對結果的影響有多少. 簡易的分析可以計算出因素與結果之間的直接關係,間接關係,個別關係,與所有因素之間的相關性等.也許你已經感覺其複雜性,但為追求真正的前因後果,多重迴歸分析在科學研究是一門相當重要的科目.複雜的分析通常包含更多的因素與不同的結果.例如,某一研究員想知道新方法的效果及被喜歡的程度.常見的分析錯誤是一項一項檢驗;先分析效果如何再分析受測者所持的態度.由於結果之間有其關聯性(也就是所謂的共同性),分別計算的方式會誇大其結論.所以演生出MANCOVA及MANOVA(若有興趣請查Google).但科學界還是建議Less is more (Cohen, et. al. 2003).意思是少一些外來因素所得的結果才會有更多的效度.科學研究不就是要人相信結果是有效可信的嗎?社會科學當然也是一樣,所以你可以發現先進國家絕對不會將不同選舉放在一起,更不用說是用手段綁大選.但似乎台灣理解這種道理的人不夠多,才會吵鬧不休無法帶領群眾走向正途.你們覺得呢?