ความแม่นยำของการประมาณค่าคุณลักษณะส่วนบุคคลโดยใช้ GGUM

Abstract

This research aimed to study the accuracy of estimated individual characteristic using the generalized graded unfolding model (GGUM). The researchers used simulated data generated by Monte Carlo sampling under the following circumstances: (1) all Item location parameter on the characteristic scale were positive only, and both positive and negative, (2) the number of items was 10, 15 and 20, and (3) the sample size was 100, 300, 500 and 1000. Consequently, there were 24 simulations, and each simulation was repeated 100 times. The accuracy of estimated characteristic was measured by the mean correlation coefficient between individual characteristic estimates from the GGUM and true individual characteristic values.

          As a result, it was found that (1) tests with both positive and negative Item location parameter possessed high accuracy of estimated individual characteristic than tests with only positive item parameters, (2) when the sample size was constant, the difference in numbers of items affected the accuracy of estimated individual characteristic since the 15- and 20-item tests tended to achieve higher accuracy than the 10-item test, and (3) when the number of items was constant, the larger sample size tended to increase the accuracy of estimated individual characteristic since the sample size of 300, 500 and 1000 provided the same level of accuracy; however, the accuracy is greater compared to the sample size of 100.