JOURNAL OF MEASUREMENT AND EVALUATION IN EDUCATION AND PSYCHOLOGY-EPOD, vol.9, no.1, pp.49-63, 2018 (ESCI)
The aim of the research was to examine the effects of the number of categories for polytomous items on psychometric properties in a nonparametric item response theory (NIRT) model. For the purpose of the study, data sets with two different sample sizes (100 and 500) that come from different sample distribution shapes (normal distribution, positively skewed distribution, and negatively skewed distribution), two different test lengths (10 items and 30 items), and three different number of categories (three, five, and seven) were generated. The effects of the number of categories on psychometric properties of polytomous items were analyzed by Mokken Homogeneity Model (MHM) under NIRT model. The research was designed as a basic research. In the generation and analysis of data sets, R Studio 3.4.0 software was used. For analysis conducted with MHM, Mokken package was used in R Studio. According to scaling with MHM, specific pattern of item fit to MHM with changing the number of categories was not observed. In general, it was found that the number of categories has no effect on reliability estimate. It was determined that tests have weak fit to MHM under test conditions in the research.