UBC Theses and Dissertations
Type I error rates of the DIF MIMIC approach using Joreskog’s covariance matrix with ML and WLS estimation Gelin, Michaela Nicole
This dissertation presents new research that examines the Type I error rate of a structural equation modeling (SEM) approach for investigating differential item functioning (DIF) in short scales. Specifically Muthen's S EM model for DIF is examined using a new estimation method (Joreskog, 2002). In general, this method conditions on the latent variable while simultaneously testing the effect of the grouping variable over-and-above the underlying latent variable of interest. Thus, it is a multiple-indicators, multiple-causes (MIMIC) model for DIF. The Type I error rates of this DIF MIMIC approach are explored using data that are reflective of short scales with ordinal item response formats typically found in the social and behavioral sciences. The variables included in this Monte Carlo simulation are 7 sample size combinations (3 equal and 4 unequal group combinations), 2 item response distributions (symmetric and positively skewed), 2 scale lengths (10 and 20 items per scale), and 2 estimation methods (maximum likelihood and weighted leastsquares). The results indicate that the Type I error rates for the DIF MIMIC model are inflated for both estimation methods under all of the design conditions. These results are discussed in the context of validity including the implications of inflated Type I error rates for items identified as displaying DIF.
Item Citations and Data