整合分析向來是管理研究上處理相關文獻量化資料一個有效的工具，由於異質的 特性，最大概似法常被建議用來估計整合分析的模式。然而，理論上，推估之一般化最小平 方法亦是一種可行的方法，卻一直受到忽略，是以，本研究主要目的即在比較這兩種估計法 在整合分析模式上的表現。藉由因子設計模擬產生資料，再利用最大概似法與推估之一般化 最小平方法進行估計，結果發現：(1)推估之一般化最小平方法在參數估計的精確性、型一 錯誤及檢定力等方面的表現優於最大概似法；(2)最大概似法在估計效率上則高於推估之一 般化最小平方法，但兩者相當接近。此外，一個實例則顯示最大概似法的估計結果可能產生 型一錯誤。除了推估之一般化最小平方法估計上的簡便性外，以上的發現意謂著：應用整合 分析的管理研究上，推估之一般化最小平方法是一個更佳的估計方法。

　　　　　In management studies, meta-analysis has been used as a powerful statistical tool to treat the quantitative data of the relevant literature. Because of heteroscendasticity, maximum likelihood method (ML) has often suggested to estimate the meta-analytic model. However, theoretically speaking, there is another method, estimated generalized least square method (EGLS). which has long been ignored. Thus, this article aims to contrast the performance of ML and EGLS on the model. Based on a factorial design, this research employs Monte Carlo simulation to generate a set of pseudo data analyzed by these two methods. The results show the following findings: (1) EGLS is superior to ML in terms of the accuracy of estimation, type one error, and testing power; (2) ML is more efficient than EGLS as far as estimation is concerned, but practically speaking, the difference is negligible. An empirical example also shows that ML may result in type one error. Therefore, in addition to the fact that EGLS is simpler for estimation, this study further proves that it is the better method of estimation for the application of meta-analysis in management studies.