Construct Validation of the English Subtests of the Korean College Scholastic Ability Test

Abstract

This study presented a construct validation procedure for the English subtest of the 1998 KCSAT (Korean College Scholastic Ability Test) for 10,000 examinees with different academic tracks (e.g., Humanities & Sciences). The English subtest consisted of two subscales (e.g., Listening and Reading Comprehension) with a total of 55 test items, and the binary item response data were subjected to the TESTFACT 4.0 program (Wood, Wilson, Gibbons, Schilling, Muraki, & Bock, 2002), which handles exploratory factor analysis with dichotomous data. Factor scores generated from the TESTFACT procedure were taken to the LISREL 8.5 program, and a multi-sample confirmatory factor analysis was performed to examine the invariance of the factor loadings of the English subtest across humanities and sciences. Results of the multi-sample analysis indicate a variant factor structure of the English subtest for the Humanities and Sciences. This suggests that the test may have different meanings for examinees with different academic backgrounds, hence threatening construct validity of the test.