在社會科學研究領域中，蒐集所得的資料往往具有「巢狀結構」(nested structure)，亦即擁有階層的特性。對於此種資料，傳統的迴歸分析常使用最基層的單位， 而忽略了隸屬於同一階層之同質性的關係，導致標準誤的誤估，而變得過小，造成迴歸係數 易達顯著；或是採較高階層為單位，將較低階層的變項合併，成為較高階層的變項之一，但 卻忽略了較低階層單位的異質性。 對於此兩難情境， 近年來所發展的階層線性模式 (hierarchical linear model，簡稱 HLM) 適合用來分析具此種特性之資料。基於此，本文 首先說明階層線性模式的二階層、三階層理論：其次，本文以教育研究的實際資料進行分析 說明，此資料涵蓋縣市、學校、個人階層；最後，本文對於線性階層模式的其他應用，包括 (一 ) 個人改變 (individual change) 的線性成長模式 (linear growth model)； (二 ) 後設分析 (meta-analysis)，做一個理論的簡述與未來發展的探討。

　　　　　The purpose of this paper is to introduce the theory of "Hierarchical Linear Model" (HLM), and discuss its development in furture research. In the study of social science, the data sets that researchers collect always have "nested structure". About this kind of data, we always use regression analysis by the lower unit or the higher unit. But, both the methods have some problems. The former will makes "standard error" too small and coeffifients will be significant, the latter will ignore variance within lower units. According to the problems, the statistics technique of "HLM" is suitable for this kind of data. First, in the paper, the author introduces theory of HLM. Second, the author analyzes data to explain its application. Finally, the author gives some promoting development about HLM.