Item Analysis of Prefecture-Wide English Tests for High School Students

Abstract

Teacher-designed English tests have mainly been analyzed from the perspective of content validity. Item response data of examinees makes it possible to analyze tests statistically. A committee of high school English teachers in Ibaraki prefecture produces two kinds of English tests (IPET) every year: Test A for first-year and Test B for second- and third-year high school students. The IPET administered in 1999 and 2000 were analyzed statistically in a previous study. This subsequent study was conducted to generalize the statistical characteristics of the tests using a larger number of tests as a base. Sixteen IPET tests (8 each of Tests A and B) administered from 1995 to 2002 were analyzed. The item difficulty and item discrimination of 760 items and the reliability coefficients of all the tests were examined. The results of the classical item analysis reveal the common characteristics of the tests. (1) Test specifications helped to generate tests with higher reliability and to avoid the creation of less discriminable items. (2) Difficult items tended to be less discriminable and items with lower discrimination indices tended to be more difficult. (3) The reliability coefficients of Test A were higher than those of Test B every year. (4) Test B had more items with lower discrimination indices. (5) When teachers attempted to design more difficult tests, they tended to create tests that were beyond the level of the students.