階層線性模式路徑圖與策略化模型建構機制

A Six-Step Standardized Model Development Strategy for Hierarchical Linear Model

多階層統計分析概念較之於傳統單一階層研究錯綜複雜，對習於單層次研究設計者而言頗難理解其中道理，又多層次分析模型建構一直缺乏策略化機制，造成無所適從的窘態。本研究提出一原發性之多階層路徑概念圖像以及六步驟策略化模型建構機制，提供初學者從事多階層資料分析之指南。文中並以美國大型國家研究計畫High School and Beyond之實徵資料，隨機抽取得160所公私立高中7,185高三學生作為分析樣本，逐步示範六步驟模式發展機制應用。研究結果顯示本研究概念圖與模型建構機制兼具探索與驗證模型策略，可有效簡化模式發展並幫助研究者建立出兼顧理論與實徵的最佳適配模式，而複核效化結果進一步顯示，二折半樣本分別驗證六步驟模式發展機制所獲得完整模式，顯示本模型建構機制所推導之完整模式具穩定性，故為一有效模型建構策略。本文最後針對可能之限制及未來發展方向提出若干具體建議。 For researchers familiar with conventional single level analysis, multilevel modelingoften appears intimidating, not only due to the novel nested structure implied by thedifferent “levels”, but also due the lack of conceptually oriented path diagrams and modeldevelopment strategies. The purpose of this paper is to remedy this situation by providinga multilevel diagramming system and developing a strategic six-step HLM modeldevelopment strategy to give researchers guidance in developing the best-fitting models. Arandom sample of data from the national study of High School and Beyond from the 80sin the US consisting of 7,185 high school seniors from 160 schools, public and Catholic,are used to illustrate the step-by-step model building strategy. Results show that a wellfittingmodel can be effectively and successfully obtained by using the proposed multileveldiagrams and the model development strategy. Results from a cross-validation study, usingtwo split-half data sets, confirm the stability of the final model. Future directions and issuesrelated to multilevel analysis are also discussed.