The Effect of Chance Success on Equalization Error in Test Equation Based on Classical Test Theory

Abstract

The aim of this study was to determine the effect of chance success on test equalization. For this purpose, artificially generated 500 and 1000 sample size data sets were synchronized using linear equalization and equal percentage equalization methods. In the data which were produced as a simulative, a total of four cases were created with no chance success, and three different levels (20%, %25, %33) of chance success and the default chance success were corrected by the correction formula. In the simulated data, four different scenarios have been created that do not include chance success and contain three different success rates (20%, 25%, 33%). Accordingly, the test equalization was performed by using linear equalization and equipercentile equalization methods under two different sample sizes and four different chance success conditions. Weighted mean square error of equating methods was found for each situation, and the method with the lowest weighted mean square error was accepted as the most suitable equating method. At the end of the study, it was found out that; while linear equating is the most suitable method for equating test points with chance success; equipercentile equating is the most suitable method for equating test points without chance success.