ERIC Number: ED333038
Record Type: RIE
Publication Date: 1991-Apr
Reference Count: N/A
The Effects of Number of Score Groups and Score Group Size on the Mantel-Haenszel Alpha.
Bradley, Raymond T.; And Others
Test data from a large licensure examination were used to investigate the impact of the number of score groups and score group size on the Mantel-Haenszel (MH) alpha statistic. Two kinds of quasi-experimental studies were conducted: (1) baseline MH analyses in which the number of score groups and the ethnicity of each randomly equivalent reference-focal group pair were held constant while score group size was systematically reduced from 50 cases to 5 cases per score group; and (2) majority group versus minority group MH analyses in which both the number of score groups and ethnicity varied as the minimum score group size was reduced from 50 cases to 10 cases per score group. The responses of 68,458 candidates to 299 multiple-choice items in a licensure examination administered in 1988 were analyzed in both studies. The results from both kinds of studies show that alpha behavior remained remarkably stable as score group size was reduced. A post hoc MH analysis extending these conditions to their logical extreme--a minimum score group size of one case and a maximum score interval of a single raw score point--also show that the alpha statistic remained stable. Fifteen tables and two graphs illustrate the analyses; and an appendix of two tables summarizes the post hoc analyses. (Author/SLD)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Identifiers: Alpha Coefficient; Mantel Haenszel Procedure; Post Hoc Methods
Note: A previous version of this paper was presented at the Annual Meeting of the National Council on Measurement in Education (Chicago, IL, April 4-6, 1991).