模式化學習型學校：多層次驗證性因素分析取向

本研究採用多層次驗證性因素分析的統計技術來模式化學習型學校的量表，分析依循Muthèn(1994)所提出的四個先行檢定步驟：對整體結構做傳統的驗證性因素分析、組問變異的估計、組內結構的估計，以及組間結構的估計。此四個步驟的結果被用來建構多層次驗證性因素分析模式。分析結果顯示在教師以及學校層次的學習型學校潛在變項結構是有所差異，就以Senge的五項修練之構面所編製的學習型校量表而言，在學校層次而言，其依然具有五個因素的結構；就教師層次而言，卻只有四個因素的結構。顯然地，學習型學校模式的建構涉及個人與群體的一種階層性系統，這種系統具有多層次的特性。在一般的教育研究中，研究者會對學校或班級抽取某些數量的問卷，這種抽樣法會破壞樣本的獨立性，而導致樣本間具有群體的效果，而當樣本間具備個體以及群體的效果時，就可能產生階層性關係。然而，傳統的統計技術往往無法處理這一類資料所產生的多層次問題，因此，採用多層次模式化的技術可以說是解決此類問題的一個相當不錯的分析策略。本研究採用多層次模式化方法用以建構學習型學校量表，分析的結果確實提升了此量表的準確性。

關鍵字

多層次驗證性因素分折；學習型學校

並列摘要

This study attempts to use the statistical techniques of multilevel confirmatory factor analysis to model the measurement of learning school. The analyses follow the four initial steps proposed by Muthèn: conventional CFA of total structure, estimation of between-group variation, estimation of within-group structure, and estimation of between-group structure. Then, the outcome of these analyses is used to specify the MCFA model. The results showed that the variation in the latent construct exist at both teacher and school levels. The constructs of five disciplines proposed by Senge showed different in these two levels. It has a five-factor structure in the school level, but four in the teacher level. Obviously, the model of learning school is a hierarchical system of individual and group. That is, it has multilevel characteristics in itself. It is none of use to use conventional statistical techniques to deal with this kind of data. Thus, the best analytical strategy is to use the technique of multilevel modeling. Using this technique to construct the measurement of learning school in this study indeed enhances the accuracy of the measurement.