Authors: Aweeporn Panthong and Pairatana Wongnam
This research aimed at studying on 2 issues: 1) Comparison of correctness of Likelihood ratio test and Bayesian procedures in the direction differential item functioning for polytomous scored items under 3 different factors; differential item functioning – DIF magnitude, difference ability distribution and sample size ratio, and 2) Comparison of power rate and type- I error rate of likelihood ratio test and Bayesian procedures under 3 different factors. Results of the study were as follows: The power rates of Maximum likelihood Estimated by Likelihood ratio test and Bayesian estimate procedures in detecting uniform Differential item functioning were equal. Proportion of DIF items in the test have effect on the Likelihood ratio test and Bayesian estimate procedures. When the ability distribution differences increased, the Bayesian estimate procedures had better control of Type I error rate inflation than other procedures. In addition, the power rates for all two procedures increased when sample sizes were increased for all reference group and focal group under almost all conditions. In general, the Type I error rates of Likelihood ratio test and Bayesian estimate procedures were within the nominal limits. The significance level of .05 was applied in all DIF analyses.
Keywords: differential item functioning (DIF), Likelihood ratio test, Bayesian procedure