Measuring growth in student achievement: Can different statistical models lead to different consequences for schools?
Date of Completion
Education, Tests and Measurements|Education, Educational Psychology
The No Child Left Behind Act of 2002 brought increased focus to state accountability systems by educators, researchers, and the general public. Accountability measured as the percent of proficient students in a given school as prescribed by the legislation presents a bleak picture of school effectiveness for schools attended by a large proportion of disadvantaged students. Thus, many researchers and some state education agencies have begun to look at different measures of growth in student achievement for school accountability. Such alternative measures suggest a more positive picture of achievement for these schools by accounting for initial group differences, student mobility, and/or other sociodemographic student characteristics. However, details on the effectiveness of different models in estimating growth in student achievement are limited. ^ In this study, eleven models of school effectiveness were applied to a common data set based on actual reading Connecticut Mastery Test (CMT) 2004 data for Year 1, and simulated data for an additional two years. Consistencies among the estimates from these models were investigated using descriptive techniques and a cluster analysis. Additionally, the relationship of each of these models to student and school demographic variables was examined to determine if certain schools were presented more favorably under certain models. The analyses demonstrated that while there were some similarities across a subset of models, different models can lead to different conclusions with regard to school effectiveness. Slope estimates of student growth over time differed greatly from the estimates under status-based approaches. Moreover, slope-based estimates had smaller correlations with school-level demographics. It was also shown that the determination of school effectiveness was not related to school-level demographics in the models that controlled for those variables. ^
Goldstein, Jessica A, "Measuring growth in student achievement: Can different statistical models lead to different consequences for schools?" (2006). Doctoral Dissertations. AAI3217033.