NotesFAQContact Us
Search Tips
ERIC Number: ED445029
Record Type: Non-Journal
Publication Date: 2000-Apr
Pages: 37
Abstractor: N/A
Reference Count: N/A
Detecting Compensatory and Noncompensatory Multidimensionality Using DIMTEST.
Deng, Hui; Ansley, Timothy N.
This study provided preliminary results about the performance of the DIMTEST statistical procedure for detecting multidimensionality with data simulated from both compensatory and noncompensatory models under a latent structure where all items in a test were influenced by the same two abilities. For the first case, data were simulated to reflect real test data in terms of descriptive statistics and classical item characteristics. In this case, DIMTEST did identify some degree of departure from essential unidimensionality for data sets from the noncompensatory model when the sample size was large and the interability correlation was low. For data simulated from the compensatory model, DIMTEST results suggested acceptance of the hypothesis of essential dimensionality. In the other three cases, data were simulated under various conditions in which the relative influence of the second dimension was greater than in Case 1. For these cases, when DIMTEST identified multidimensionality, the power increased when test length and sample size increased, and when interability correlation decreased. A question that remains unanswered is whether there are monotonic relationships between the power of DIMTEST and the degree of relative magnitude or variability for the two discrimination vectors. (Contains 8 tables and 11 references.) (SLD)
Publication Type: Reports - Research; Speeches/Meeting Papers
Education Level: N/A
Audience: N/A
Language: English
Sponsor: N/A
Authoring Institution: N/A