二分依變項相關性研究的後設分析：以留級與超齡對學童中輟的影響為例

Doing Meta-Analysis with Observational Studies with Dichotomous Dependent Variable: An Exemplary Analysis on the Effects of Grade Retention and Over-age on Students’ Dropping out of School

本研究以一個研究實例說明，從事二分依變項相關性研究的後設分析時，應注意的混淆變項問題，並提出可行的作法。文中作為範例的後設分析以探究留級經驗及超齡對於學童日後輟學的可能影響為目的，蒐集了使用21個獨立資料源的34篇文章，並登錄了66個分析模型的資料。最後，我們使用10個與14個資料源進行留級與超齡對中輟影響的後設分析。研究結果發現，有留級經驗的學童其輟學的風險約為無留級經驗學童的3倍（OR＝3.02， z＝8.36， p＜ .001），但是超齡對於中輟的影響則僅達邊際顯著（OR＝1.33， z＝1.69， p＞ .09）。研究結果很清楚地顯示，留級經驗對於學童的教育取得可能是不利的，並不支持留級可提高學童成熟度與預備度的說法。我們最後討論了此研究的意義與限制，並使用一個檢核表檢視本研究的品質。這個研究除了提供有意後設分析二分依變項相關研究的學者可以參考的範例，也說明了教育部門從事留級相關決策時應深思的訊息。

By presenting an illustrative meta-analysis， we discussed some issues to be noted when meta-analyzing results of observational studies in which dependent variables are dichotomous in nature. The purpose of the exemplary meta-analysis is to explore the possible effects of children’s retention experiences on their dropout behavior later on. We retrieved 34 articles which consist of 21 independent data sources and coded 66 effect-size indices， out of which 10 were used for analyzing the effects of retention and 14 were used for analyzing the effects of over-aging.The findings are that， while students with retention experiences were three times more likely to drop out of high school than those without (OR＝3.02， z＝8.36， p＜ .001)， the effects of over-aging reached only marginal significance (OR＝1.33， z＝1.69， p＞ .09). The results suggest strongly that being held back by the school could have significant negative impacts on students and the practice of retention did not help students to be more mature and ready for school learning. After some discussion on the implications and limitations of this meta-analysis， we use a check list to examine the merit of this meta-analysis.