Comparison among cognitive diagnostic models and item response models in a Japanese sample of TIMSS 2007 mathematics data

Abstract

Although Japan has been participating in the Trends in International Mathematics and Science Study (TIMSS) since its inception, it is still not known what kind of cognitive diagnostic models best describes the Japanese sample. Therefore, in this study, we compare the relative fit of a variety of cognitive diagnostic models in the Japanese sample of the TIMSS assessment. We also investigate the possible reasons for the differences in model fit. The results of the comparison and investigation are as follows. The reduced reparameterized unified model (R-RUM) was found to be the best in terms of both AIC and BIC. The good fit of the R-RUM can be attributed to the many cognitive elements required to solve the TIMSS items, as well as their interactions with each other. In addition, the diagnostic models fitted better to the data than the item response theory models.