Please use this identifier to cite or link to this item: https://ah.lib.nccu.edu.tw/handle/140.119/97622
題名: 學校效率研究中的因果推論:以階層線性模式為例
作者: 王文中
關鍵詞: 學校效率 ; 階層線性模式
日期: Sep-1995
上傳時間: 4-Jun-2016
摘要: 在Suppes的因果機率理論中,「真因」的探索是個哲學上問題,任何統計方法都 幫不上忙。統計方法所能做的就是,假定那個事件是因之後(不管它是否就是真因),估計 它所帶來的效果有多人。因此,問題的重心不在於尋找效果的因( cause of an effect ) ,而是估計因的效果( effects of a cause )在本文中,我介紹 Rubin 的實驗模式。 他 的模式幫助我們澄清因果推論的幾個重要問題。例如透過隨機分派的實驗,從實際資料中所 獲得的實驗處理的參數,才不會只是關聯參數( associational parameters ),而可以成 為因果參數( causal parameters )。 在觀察的研究中,如果發現各組的共變項顯著不同 , 那麼必須假設「強忽略性」( strong ignorability ),從資料中求得的參數才會有因 果上的意義。我以學校效率的研究為例,說明如果要講這類研究所獲得的參數具有因果推論 的意義的話,所必須檢驗的假設。由於這類的研究常以階層線性式來估計參數,所以我先大 致介紹了階層線性模式的理論基礎和參數估計方法,並以三個例子說明如果這些假設成立的 話,哪些參數可以被用來當作學校效率的指標。
According to Suppes`s probabilistic theory of causality, search for genuine cause is a philosophical issue. None of statistical methods can help determine the genuine cause. What statistical methods can do is to estimate effects of a presumed cause, no matter it is a genuine cause or not. Therefore, we should not focus on search of a cause of an effect, but on effects of a cause. I briefly address Rubin`s experimental model to help clarify several major issues in causal inference. For instance, only through random assignments can esimated parameters possess causal meanings. Similarly, in observational studies, strong ignorability should be checked if parameters are to have causal meanings. I illustrate how to apply Rubin`s model to research of school effectiveness through three examples. Since hierarchial linear models are commonly used in this area, I concisely introduce their theoretical grouds and parameter estimation procedures. If the necessary assumptions are sustind, parameters of hierarchical linear models can be used to depict school effectiveness.
關聯: 教育與心理研究, 18,51-81
Journal of Education & Psychology
資料類型: article
Appears in Collections:期刊論文

Files in This Item:
File Description SizeFormat
index.html115 BHTML2View/Open
index.html151 BHTML2View/Open
Show full item record

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.