The Relation of Item Difficulty Between Classical Test Theory and Item Response Theory: Computerized Adaptive Test Perspective

Abstract

This study aims to transform the calculated item difficulty statistics according to Classical Test Theory (CTT) into the item difficulty parameter of Item Response Theory (IRT) by utilizing the normal distribution curve and to analyze the effectiveness of this transformation based on Rasch model. In this regard, 36 different data sets created with catR package were studied. For each data set, item difficulty parameters and transformed item difficulty parameters were calculated and the correlation coefficients between these parameters were analyzed. Then, Computerized Adaptive Test (CAT) simulations were performed using these parameters. According to the simulation results, the correlation coefficients between the estimated theta values with both methods were high. Furthermore, in CAT simulations in which both parameters were used, especially in the samples which were over 250, it was found to have similar bias, RMSE values, and the average number of administered items.