The Performance of Dimtest When Latent Trait and Item Difficulty Distributions Differ

Publisher Website
Google Scholar
Abstract
This monte carlo study examined DIMTEST for unidimensional and two-dimensional data that exhibited ceiling effects. These effects were generated through changes in the location and scale of the θ distribution when the item difficulty (b) distribution was held constant at b ~ N(0,1). Results indicated that the power of DIMTEST was reduced as the location shifted upward and the scale shifted downward. This held only when the AT1 items were selected using principal axis factor analysis. When AT1 items were selected using expert opinion, the procedure was only slightly affected by changes in location and scale. However, it is unclear how successful the use of expert judgment will be in applied settings.

This publication has 31 references indexed in Scilit: