Measurement invariance is a critical issue in psychological and educational research because it is often required for meaningful group comparisons. A designated, unbiased anchor is desired for any invariance analysis, because the bias in an innocuously chosen anchor may lead to erroneous conclusions about item-level invariance, or equivalently, differential item functioning (DIF). This article presents an application of the effects-coded scaling as an alternative to the conventional, marker-variable scaling in mean and covariance structure (MACS) DIF analysis. By means of Monte Carlo simulation, this article examined if this new approach possibly leverages the problems of a biased anchor. Simulation results suggest that the effects-coded method provides some favorable outcomes over the marker-variable method but more research is needed to refine MACS DIF analysis.
|Journal||Journal of Educational Evaluation|
|State||Published - 2013|