Item Analysis of Criterion-Referenced Tests

Abstract

We propose that teachers should not calculate reliability but rather should use item analysis data as the major determinant of the adequacy of classroom measures. Using the results of two actual classroom tests-one elementary-level mastery test and one secondary-level nonmastery test–we demonstrate an approach to item analysis for CRTs that involves the calculation of item difficulty for all items and discriminating power for those items with unacceptable P values. For each test, the effects on initial reliability estimates and item analysis indexes of systematically removing low-scoring students’ results (one at a time) are presented. We conclude that this modified item analysis strategy for classroom CRTs is both efficient and useful.