Beschreibung:
<jats:p> Latent trait procedures are being proposed for use in the construction and scaling of standardized achievement tests. These procedures, however, usually require rather strong assumptions that will not be met in any absolute way for data from such tests. The primary purpose of this study was to compare the relative robustness of the one-, two- and modified two-parameter latent trait logistic models for one specific standardized test. Responses of examinees in grades 9 (N = 944) and 12 (N = 650) to five subtests of the Iowa Tests of Educational Development were analyzed. The one-parameter model misfit substantially more items than either of the two models (% misfit 28-63% vs. at most 39%). The items identified as misfitting by the modified two-parameter model were slightly more uniformly distributed over a test's table of specifications than were the misfitting items identified by the other two models. These results suggest that of the three models studied, the modified two-parameter model may provide the best representation of the data from tests such as the ITED. </jats:p>