Accuracy of Measurement in the Classical and the Modern Test Theory: An Empirical Study on a Children Intelligence Test

The study aimed to compare the accuracy of assessing participants’ ability by the significance of standard error of the Classical Test Theory (CTT) and standard error of estimation of the Modern Test Theory (MTT) represented by the Two-Parameter Logistic Model (2PLM). It also aimed to compare item difficulty and arrangement in the two theories using Attriri’s Intelligence Scale for Children and a sample of 2674 students from the Republic of Yemen. Descriptive statistics (means and standard deviations), exploratory factor analysis and one-sample t-test were used for statistical treatment of data. Statistical treatment was performed by the IBM SPSS V. 20 and the Bilog-Mg3 programs. It was found that MTT represented by the 2PLM is more accurate than CTT in assessing participants’ abilities by standard error. Furthermore, the calibration of items by difficulty and the arrangement of participants’ abilities in the two theories proved to be different. Based on the study results, the researcher recommends (a) basing the development of psychological tests on the psychometric characteristics extracted according to MTT, (b) training professionals in measurement and evaluation in the use of analysis programs to extract item and ability parameters according to the different models of MTT (item response theory), and (c) making available the programs needed for the use of MTT in testing, e.g., Xcalibre and Rumm 2030 & Bilog-Mg3.