Evaluating measurement equivalence is a necessary first step before comparisons can be made across groups or over time. As a result, techniques for evaluating equivalence have received much attention in the literature. Given the many benefits of these approaches, measurement equivalence is most appropriately assessed using item response theory (IRT) or confirmatory factor analytic (CFA) techniques. For both methods, the identification of biased items typically involves statistical significance tests based on the chi-square distribution or empirically derived rules of thumb for determining nonequivalence. However, because of the disadvantages of these criteria, it may be informative to use effect size estimates to judge the magnitude of the observed effects as well. As such, the present work proposed the development and evaluation of effect size measures for CFA and IRT studies of measurement equivalence. First, simulation research was used to illustrate the advantages of effect size measures and to develop guidelines for interpreting the magnitude of an effect. Next, these indices were used to evaluate nonequivalence in both cognitive and noncognitive data. In sum, the results show the benefits of evaluating the effect size of DIF in addition to assessing its statistical significance.
【 预 览 】
附件列表
Files
Size
Format
View
The development and validation of effect size measures for IRT and CFA studies of measurement equivalence