ERIC Number: ED212649
Record Type: Non-Journal
Publication Date: 1981-Nov
Reference Count: N/A
Evaluation Design Project: Multilevel Interpretation of Evaluation Data Study.
Miller, M. David; Burstein, Leigh
Two studies are presented in this report. The first is titled "Empirical Studies of Multilevel Approaches to Test Development and Interpretation: Measuring Between-Group Differences in Instruction." Because of a belief that schooling does affect student achievement, researchers have questioned the empirical and measurement techniques used to evaluate the effects of schooling on student achievement. One possible shortcoming of the major standardized norm-referenced achievement tests is their failure to take into account the nesting of units in the educational system. This study uses item data at the class level to construct tests. It was found that selecting items from an index of discrimination between groups did lead to scales more sensitive to instructional differences. From the analysis of patterns of item response, it was found that not only did patterns of correct and incorrect item response vary as a function of class membership, but that the patterns of response reflect substantively meaningful differences in instruction. The second study, "State of the Art Methodology for the Design and Analysis of Future Large Scale Evaluations: A Selective Examination," reviews how recent methodological advances might be incorporated in future large-scale evaluations. Specifically, structural equation modeling and selection modeling and related issues in analysis of quasi-experimental data are examined. (Author/BW)
Publication Type: Reports - Research
Education Level: N/A
Sponsor: National Inst. of Education (ED), Washington, DC.
Authoring Institution: California Univ., Los Angeles. Center for the Study of Evaluation.