- 作者: 高新建
- 作者服務機構: 臺北市立師範學院初等教育學系
- 中文摘要: 教育活動通常發生在具有階層特性之「內屬結構」的教育情境;相對應地,教育實證研究通常也有描述不同層次的資料。由於統計技術及電腦應用程式在近十餘年來的進步,最近所發展出來的「階層線性模式」,使得教育研究人員能夠將影響教育成效的因素是來自多個層次的事實,納入其研究模式,以計算各個層次之預測變項的效果,並克服傳統線性模式研究常會遭遇到的三項問題,進而能夠精確地估計出標準誤、正確地檢驗出迴歸的異質性、並避免合計的偏差。本文採用「階層線性模式」的概念及其統計程式,分析從美國「1988年全國教育縱貫研究」公用版光碟片資料庫所抽取樣本資料,以說明階層線性模式的基本模式,並與「散計」及「合計」二種傳統一個層次的迴歸分析方式作比較,以展現階層線性模式的特色與優點。
- 英文摘要: Education usually occurs in a context featured with the nested hierarchical structure. Empirical studiesin the field of education, correspondingly, tend to have data delineating each level of the hierarchy. Due to thedevelopment of estimation techniques and of statistical computing programs in the past decade, the newlydeveloped "hierarchical linear models"(HLM) enables researchers to model the phenomenon that educationis influenced by factors resided in a multilevel structure and to calculate the effects that occur at each of thelevels. These advances help resolve the three most commonly encountered difficulties of past analyses ofmultilevel data: misestimated standard errors, heterogeneity of regression, and aggregation bias. Applyingconcepts and computing program of HLM to analyze a subsample drawn from the "National EducationLongitudinal Study of 1988" public use CD-ROM data base, this study presents basic models, concepts,features, and advantages of HLM. These characteristics are demonstrated by a comparison between the HLMapproach and two commonly used regression analysis techniques, disaggregation and aggregation.
- 中文關鍵字: 內屬結構資料; 多層次分析; 階層線性模式; 學習機會
- 英文關鍵字: hierarchical linear models; multilevel analysis; nested data; opportunity to learn