Comparing item-total correlation and item-theta correlation in test item selection: A simulation and empirical study

Abstract

One of the important processes in the evaluation of the psychometric properties of a test is item selection. The item selection process usually uses a very popular technique called item-total correlation. This study attempts to describe the item-total correlation technique and explore it using a similar technique called item-theta correlation. Both techniques are applied using simulation studies by creating several conditions related to test length and sample size. After the simulation study, the next step is the study using empirical data as an illustration of the results of the simulation study. The results of this study show that there are differences in the results of item selection based on these two approaches. Item-theta correlation detects more items that have weak discrimination power than item-total correlation. The difference is more noticeable in conditions where the cutoff point used for item selection is low(.20).