ERIC Number: ED333030
Record Type: RIE
Publication Date: 1991-Apr
Reference Count: N/A
Assessing Dimensionality of a Set of Items--Comparison of Different Approaches.
Performance in assessing the unidimensionality of tests was examined for four methods: (1) W. F. Stout's procedure (1987); (2) the approach of P. W. Holland and P. R. Rosenbaum (1986); (3) linear factor analysis; and (4) non-linear factor analysis. Each method was examined and compared with the others using simulated and real test data. Seven data sets were simulated, three unidimensional and four two-dimensional, all with 2,000 examinees. Two levels of correlation between abilities were considered. Eight different real test data sets were used: four were unidimensional, and four were two-dimensional. Real data came from the National Assessment of Educational Progress tests for U.S. history and literature for grade 11 and from the Armed Services Vocational Aptitude Battery for grade 10 for arithmetic reasoning and general science. Findings suggest that, while linear factor analysis overestimated the number of underlying dimensions, the other three methods correctly confirmed unidimensionality but differed in their ability to detect a lack of dimensionality. Stout's procedure showed excellent power in detecting a lack of unidimensionality. Holland and Rosenbaum's procedure and the non-linear factor analysis approach showed good power provided the correlation between abilities was low. A 40-item list of references is included. Four tables present study data. (SLD)
Publication Type: Reports - Evaluative; Speeches/Meeting Papers
Education Level: N/A
Authoring Institution: N/A
Identifiers: Armed Services Vocational Aptitude Battery; National Assessment of Educational Progress; Unidimensionality (Tests)
Note: Paper presented at the Annual Meeting of the American Educational Research Association (Chicago, IL, April 3-7, 1991).