Type I Error Rates and Power Study of Several Differential Item Functioning Determination Methods

Abstract

In this study Type I Error rates and power study of several differential item functioning determination methods were investigated for several conditions. For this purpose simulation studies were conducted for methods which belong to Classical Test Theory (CTT) and Item Response Theory (IRT). The results for Type I Error rates of differential item functioning determination methods under the framework of CTT show that none of manipulation criteria have significant effect on average Type I Error rates. For IRT based
methods it has been observed that the increase of sample size causes the decrease of average Type I error rates while increasing the number of items has no significant effect on average Type I Error rates, and that the equal ratio of reference and focal sample sizes cause lower average Type I error rates. The results for power study for both differential item functioning determination methods of CTT and IRT show that the increase of sample size and the magnitude of differential item functioning cause the increase of average power rates. No significant effect of the number of items on average power rates could be observed.