The Prediction of Test Realibility and Test Information Function According to Classical Test Theory and Latent Trait Theory

Abstract

In this study, the prediction power of the item characteristics based on the experts’ predictions in exams which cannot be piloted was examined for test reliability and test information functions. For this aim, classical test theory and logistic model with two parameters of latent trait theory were used. The study was carried out on the results of the math sub-tests of 9914 randomly selected students who took The Student Selection and Placement Examination for Secondary' Schools in the 1999-2000 academic year. The experts’ prediction values were obtained from separate metric scales based on the above mentioned theories. These scales were given to 16 math teachers.The results of the study revealed no significant difference (0,05) between test reliability values obtained through experts’ predictions and through classical test theory. Furthermore, test information functions taking place in latent trait theory was examined within the scope of this study. The findings showed that test information function values predicted for different competence levels increase for individuals having -2 ve +0,5.competence level, test information function values decrease for the individuals having +0,5 and +2.