Empirical Demonstration of Techniques for Computing the Discrimination Power of a Dichotomous Item Response Test

Abstract

The study was designed to examine the various statistical techniques for computing the Discriminating power of a dichotomous item response test. The simple Discrimination Index, the Cramer’s phi coefficient and conventional phi coefficient, Point Biserial Correlation, and t-test for independent sample was used to test the difference in mean of the criterion scores between those who got the item correct and wrong. Item 1 of the Senior Secondary Agricultural Science test was used as the computational example. In the analysis, the simple Discrimination Index and 3 formulae of the Point-Biserial correlation gave the same result (r = 0.63). The Cramer’s phi coefficient was 0.49 and phi coefficient (0.65). There was significance difference between the mean criterion scores of students who got the item right and those who failed the item (t+ 5.05 p < 0.05).The phi coefficient, point-Biserial correlation and t-test for independent sample were the best statistics for estimating the Discrimination power and coefficient of the test item. The simple Discrimination Index did not take into cognisance the performance of candidates in the test/. The phi coefficient is recommended where the sample is large and necessitates division into upper and lower extreme groups.

DOI: 10.5901/jesr.2014.v4n1p189