ERIC Number: ED213729
Record Type: RIE
Publication Date: 1979-Nov
Reference Count: 0
Multilevel Approaches to the Evaluation of Educational Effectiveness. Studies in Measurement and Methodology, Work Unit 2: Analysis of Multilevel Data.
Conklin, Jonathan E.; Burstein, Leigh
Educational outcomes are affected by student level, classroom level, and school level characteristics. The fact that educational data are multilevel in nature poses serious analysis questions. Though strong arguments can be made for focusing on a single level of analysis, such studies have several basic limitations: the choice of analytic level should be based on substantive, not statistical, concerns; inferences may be affected by aggregation bias; analyses at one level may automatically alternate the effects of variables from other levels; and they reveal little information about processes that exist within classrooms or within schools. This paper reviews evidence to support the claims that: distinct and important issues are addressed at each analytical level, that relationships may change at different levels, and that these changes may provide information not obtainable in any single level of analysis. Two studies are reviewed which empirically examine these issues and analyses are performed on an available database to provide specific examples of effects at multiple levels. (Author/BW)
Publication Type: Reports - Research
Education Level: N/A
Sponsor: National Inst. of Education (DHEW), Washington, DC.
Authoring Institution: California Univ., Los Angeles. Center for the Study of Evaluation.
Identifiers: Generalizability Theory; Unit of Analysis Problems
Note: Paper presented at the Annual Meeting of the American Educational Research Association (64th, Boston, MA, April, 1980).