國家發展、學校因素及學生因素對數學成就影響之多層次分析

A Multi-Level Analysis of the Influence of National Development and School and Student-level Factors on Mathematics Achievement

本研究以38個國家參與TIMSS 2015年資料分為國家、學校與學生層因素， 透過多層次模式以6,375所學校， 206,002名八年級生分析獲得結論如下：一、國家、學校與學生之間差異在學習成就解釋力各有34.96%、23.0%及42.04%。二、學生的家庭資源、學生歸屬感、學生霸凌、學習數學信心及學生認知的數學價值對數學學習成就都是重要影響因素。三、學生來自富裕家庭比率愈高、學校愈重視學業表現及學校紀律問題愈少，數學學習成就愈好。四、國家的國民所得、性別平等、國家競爭力、教育年數對學習成就都是重要影響因素。本研究貢獻在於以三層模式分析各國八年級學生數學成就表現因素發現，跨國與校際差異對學習成就具有重要解釋力，以及各層納入的變項都是重要因素。

This study developed a multi-level model comprising nation, school, and student-level factors utilizing data from 6,375 schools and 206,002 eighth-grade students from 38 countries participating in the 2015 TIMSS international assessment survey. Analysis of the model lead to the following conclusions: 1. The explanatory power, in terms of differences in learning achievement, was 34.96% for country-level factors, 23.0% for school-level factors, and 42.04% for student-level factors. 2. Among student-level factors, family resources, student sense of belonging, the degree of bullying, students’ confidence in learning mathematics, and the value students place on mathematics were important factors influencing mathematics achievement. 3. In terms of school-level variables, higher achievement in mathematics was associated with schools having a higher proportion of wealthy parents, placing greater emphasis on academic performance, and reporting fewer discipline problems. 4. National-level factors influencing learning achievement included national income, gender equality, national competitiveness, and the number of years of compulsory education. The contribution of this research is in the application of a three-level model to analyze the factors influencing mathematics achievement for eighth grade students among 38 countries, finding that transnational and inter-school differences have important explanatory power in terms of learning achievement, with specific variables included at each level (including country, school, and student) serving as important factors.