有鑑於傳統的整合性分析只考慮單一層次和效應量的集中趨勢，本研究提出多層次的整合性分析，同時探應效應量的集中和分散情形。多層次模式較能貼切資料的結構，避免因為模式錯誤所產生的統計偏誤。而分析效應量的分?情形，可以更具體看出影響效應量分?的原因和預測因子，除有助整合性分析外，也可提供後續實驗操弄的建議。我們介紹了適用於整合性分析的多層次模式，以及參數估計方法和意義。最後透過簡單例子，呈現此一分析方法的應用的意涵。

　　　　　Conventional meta-analysis is uni-level and focuses on central tendency of effect sizes. In this study, we propose multi-level meta-analysis that can explore both central tendency and variability of effect sizes. Multi-level analysis better reflects hierarchy of data structures and thus avoids statistical bias due to model mis-specification. Analysis of variability of effect sizes better depicts influencing factors on the variability, which not only is helpful to meta-analysis but also gives suggestions for further research design. We propose several sub-models for meta-analysis and explain estimation procedures and meanings of the parameters. Through real data analysis, implications and applications of the proposed analysis are addressed.