ERIC Number: ED397100
Record Type: RIE
Publication Date: 1996-Apr
Reference Count: N/A
Assessing the Dimensionality of Item Response Matrices Using a Goodness-of-Fit Index Based on Noncentrality.
De Champlain, Andre
The usefulness of a goodness-of-fit index proposed by R. P. McDonald (1989) was investigated with regard to assessing the dimensionality of item response matrices. The m subscript k index, which is based on an estimate of the noncentrality parameter of the noncentral chi-square distribution, possesses several advantages over traditional tests of hypotheses as well as other descriptive fit indices. This study considered the behavior of the index in simulated conditions across different test lengths, sample sizes, dimensional structures, and other factors. The appropriateness of the recommended model fit cutoff value (0.9) was also studied. Four hundred unidimensional data sets based on the Law School Admission Test were simulated for the study. Results suggest that the index cutoff value recommended by its developers as being indicative of model fit is too high for the data sets simulated. With respect to the simulated unidimensional data sets, results show that none of the manipulated factors had any practical effect on mean m subscript k index values, supporting claims that the index is sample-size and estimation independent. These findings may offer some guidelines to the practitioner interested in using the index to assess the dimensionality of an item response matrix. (Contains 1 figure, 8 tables, and 46 references.) (SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Identifiers: Dimensionality (Tests); Law School Admission Test
Note: Paper presented at the Annual Meeting of the American Educational Research Association (New York, NY, April 8-12, 1996).