Investigating the correlations between response time, ability level and item parameters in computer adaptive test

This research was conducted to examine the relationship between item response time, ability level and item parameters in Computerized Adaptive Test (CAT). For this purpose, real-time CAT was prepared with the 2017 autumn, 2018-1/2/3 semester questions of the Academic Personnel and Graduate Education Entrance Examination (ALES). Item and ability parameters were obtained from the candidates who took the exam during these periods using the Rasch Model, one of the item response theory (IRT) models. 238 individuals completed the numerical test and 203 individuals completed the verbal test of the CAT prepared with these item parameters. CAT was applied to senior undergraduate students and undergraduate graduates who were eligible to take ALES. During the CAT, the time individuals spent to answer each item was recorded. Lognormal item response time model (LNIRT) analyses, in which the time spent answering items is included in item and ability estimation, were conducted in R Studio with the help of the LNIRT package. The relationship between item and ability parameters obtained with LNIRT and item and ability parameters obtained with IRT's Rasch Model and 2 parameters logistic model (2PLM) was examined. Although the relationship between the item parameters obtained with the IRT, which does not consider the response time, and the item parameters obtained with the LNIRT varies according to the IRT model (Rasch-2PLM), a significant relationship was found between the time parameters and the item parameters. There was a negative relationship between the speed parameter estimated by LNIRT and ability values. It was concluded that the time spent on the item and the total test was effective, especially for individuals in low and high ability groups. Additionally, a relationship was obtained between the ability and speed values obtained from the LNIRT and MTK models. Keywords: computerized adaptive test, item response theory, response time, lognormal response time item response theory model