ERIC Number: ED221558
Record Type: Non-Journal
Publication Date: 1981
Reference Count: N/A
Empirical Studies of Multilevel Approaches to Test Development and Interpretation.
Burstein, Leigh; Miller, M. David
Instructional and program relevance of measures of student performance and use of standardized achievement tests, as measures of the quality of a student's educational experiences, are discussed. This focus is consonant with current emphasis on linking testing and instruction and on systemic efforts at program and instructional improvement. Psychometric and multilevel educational data considerations are of particular concern. Two general approaches to investigation of between-group (classroom, school, program) differences in test performance can be used to detect differences in content, sequence and quality of instruction. Patterns of data responses from a standardized testing program conducted within a school district in the Beginning Teacher Evaluation Study (BTES) were examined. Between-group indices of the distribution of performance by class, school and ethnic background were generated under alternate rules for content classification. The subsets of group sensitive items and between-class correlation of items with instructional variables were used to build a scale sensitive to the variable used to select items. Item response patterns were related to group membership and instructional coverage. The results supported the argument that tests can be constructed in multilevel approaches sensitive to desired group characteristics for information about instructional experience differences. (CM)
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.