分析單位改變對跨國比較研究的影響—以TIMSS數學學習心理特質與學習成就之關聯性探究為例   全文下載

The Influence of Changing the Unit of Analysis on A Cross-National Comparative Research: Using TIMSS Databases to Examine the Relationships Between Student Psychological Characteristics in Mathematics Learning and Achievement

國際教育調查研究常採用巢套模式蒐集資料，資料具有多層級之性質，惟當資料的分析單位愈多元，愈容易產生跨層級邏輯推論的謬誤，認為個體的性質在群體中也會出現，或群體的特質可推論到個體層次，此類現象泛稱為生態謬誤。國外研究指出，生態謬誤對於教育政策評估、規劃與資源分配影響甚巨，惟國內教育研究對此議題尚未充分討論，故本研究以國際學力評鑑報告書及文獻為基礎，探究生態謬誤之現象，並利用次級資料分析，檢驗TIMSS資料庫中分析單位的轉換對於學生數學學習心理特質與成就關聯性的影響。研究發現，同層級分析與聚合分析結果相異，驗證了分析單位的轉換會影響研究結果的論述，且生態謬誤具有跨時空的特性，據此本研究研提有關分析單位選擇及資料詮釋的建議，以期減少生態謬誤之發生。

Cross-national assessments and research projects on student learning performance tend to utilize nested model to collect data. As data come from multiple levels, the unit of analysis becomes diverse and increases the difficulty on data interpretation, even results in ecological fallacy. The fallacy refers to improperly infer findings from individual-level data to group-level and vice versa. Such fallacy will affect the accuracy of educational policy analysis, planning and resource allocation. Literature review showed limited educational research in Taiwan related to this issue. Therefore, the study conducted a secondary data analysis to test the fallacy by examining the relationship between student learning characteristics and achievement in mathematics. The data were derived from TIMSS 2003 and 2007 databases. The findings indicated that student-level results were different from those of national level, leading to a conclusion that the ecological fallacy occurred as the unit of analysis changed. Moreover, the fallacy was observed across countries and times. Finally, recommendations were discussed for data interpretation in future study.