調查數據的事實與反事實研究：以商管學生就學滿意度為例   免費試閱

Counterfactual Analysis of the Survey Data: An Example of Student Evaluation of Schooling Satisfaction

因果推論不僅是科學研究的程序，也是其目的，但是實驗典範的因果檢驗技術並未能充分論斷因果關係，反事實分析則是基於實驗研究中的「因」與「果」的重新定義，發展出一套用來檢驗因果關係的實證分析模型。此一分析模型除了應用在實驗研究之外，亦可應用於調查研究所得到的觀測資料分析。本研究除了概述反事實分析的基本原理與統計檢驗程序之外，並以臺灣高等教育資料庫的4,740位商學院學生調查資料，以「入學方式」為分組變數，分析學生在「學校滿意」、「系所滿意」、「學習滿意」三個就學滿意度變數上的潛在結果，進行事實與反事實數據的比較分析。分析結果發現，不同入學方式的學生在三個滿意度變數上的差異，經過反事實分析調整後差異縮小，選擇偏誤的修正改變了分組變數的組間差異。敏感度分析結果指出，反事實分析降低組內變異的異質性，維持殘差常態性，並降低影響值對殘差估計的影響。本研究從實徵數據與敏感度分析檢驗了反事實分析的特性，也討論反事實分析的一些限制與未來研究建議。反事實分析的方法學意義對於社會科學研究資料分析具有重要的啟示，值得廣泛且深入討論。

Causal inference is not only the process, but also the purpose of the scientific research. However, the examination of causal relationship in the traditional experimental paradigm would not fully assure the causality of variables. In taking a critical viewpoint, the counterfactual analysis redefine the “cause” and “effect”in a study, proposing an empirical procedure to examine the causal relationship in experimental as well as non-experimental research, such as a survey. In the present study, the basic principles of the counterfactual analysis were reviewed. An empirical example based on a total of 4740 third-grade students of the business schools selected from the Taiwan Integrated Postsecondary Education Dataset demonstrated the application of counterfactual analysis of the survey data of learning satisfaction. “Enrolling type”was taking as the treatment of the cause of the three outcomes of “school satisfaction”, “department satisfaction”, and “studying satisfaction”. Results revealed that the counterfactual analysis successfully taken the selection bias into account. Positive evidences of sensitivity analysis also supported the superiority of the counterfactual analysis. The advantages of this new approach to survey data were discussed. The methodological insights were also clarified in the paper.