Item Analysis of a Multiple-Choice Rational Cloze Test

概要

論文の詳細を見る
The basic unit of any test is the test item. In order to make a test better, we need to identify weak items and modify them. The purpose of this investigation is to identify and compare the weak items of a cloze test by classical test theory (CTT) and item response theory (IRT) measurement. A four-option multiple-choice (MC) rational cloze test was developed. According to the range of context for closures, words deleted can be categorised into two types: higher-order and lower-order items. Item difficulty, item discriminability with sample separation, and point-biserial correlation were used for CTT analysis. As for IRT analysis, PROX was employed for the Rasch one-parameter calibration of test items. We found that some items were accepted by CTT but rejected by IRT method, and others were judged as weak by CTT but considered to be acceptable by IRT approach. In order to ascertain whether item difficulty will be ordered according to level of context required for closure, higher-order and lower-order items were compared. The result shows that higher-order items are not always more dificult than lower-order items. It is claimed that a MC format encourages wild guessing so that difficulty level of each item is determined by what options are given. We also noted that item difficulty correlated moderately with item misfit. This result implies that random guessing increases with increasing difficulty.
桃山学院大学の論文
1998-12-21