Comparison of reliability and information concepts of classical and moderntesttheory with polytomous item tests

Data obtained from psychological measures always include some error varianc e. The score obtained from one administration is only one of the observable values from its possible score distribution. Reliability and standard erro r of measurement enable test users to estimate the confidence interval whic h contains the person's true score. In modern test theory, the values of st andard error of measurement provide different values for individuals who ar e at different regions on the measured latent trait. Furthermore, item and test information functions in modern test developing models offer an altern ative way of constructing more reliable psychological tests In this study, a personality scale which has polytomous items, analyzed with both the Clas sical Test Theory and Item-Response Theory (IRT). It can be concluded that IRT concepts, such as item and test information functions, seem to be more efficient than the classical item parameters in making more precise distinc tions between different regions of the latent trait.