There are two serious methodological problems in the research literature on school effectiveness, the ecological problem in the analysis of aggregate data and the problem of not controlling for important confounding variables. This dissertation corrects these errors by using multilevel modeling procedures, specifically Hierarchical Linear Modeling (HLM), and the Canadian Trends in International Mathematics and Science Study (TIMSS) 2007 data, to evaluate the effect of school variables on the students’ academic achievement when a number of theoretically-relevant student variables have been controlled. In this study, I demonstrate that an aggregate analysis gives the most biased results of the schools’ impact on the students’ academic achievement. I also show that a disaggretate analysis gives better results, but HLM gives the most accurate estimates using this nested data set.
Using HLM, I show that the physical resources of schools, which have been evaluated by school principals and classroom teachers, actually have no positive impact on the students’ academic achievement. The results imply that the physical resources are important, but an excessive improvement in the physical conditions of schools is unlikely to improve the students’ achievement. Most of the findings in this study are consistent with the best research literature. I conclude the dissertation by suggesting that aggregate analysis should not be used to infer relationships for individual students. Rather, multilevel analysis should be used whenever possible.
| Identifer | oai:union.ndltd.org:MANITOBA/oai:mspace.lib.umanitoba.ca:1993/9590 |
| Date | 18 October 2012 |
| Creators | Wei, Yichun |
| Contributors | Clifton, Rodney (Education), Renaud, Robert (Education) Roberts, Lance (Sociology) Metge, Colleen (Pharmacy) Ma, Xin (University of Kentucky) |
| Source Sets | University of Manitoba Canada |
| Detected Language | English |
Page generated in 0.0022 seconds