Estimating the Selection Probability of Multiple-Choice Questions from the Similarity of Answer Choices

Shinichi Ikeda, Teruhiko Takagi, Masanori Takagi, and Yoshimi Teshigawara

Keywords

Item Response Theory, Etesting, Difficulty Level, Similar Questions

Abstract

In recent years, web-based testing, or "e-testing," has been attracting much attention. In general, item response theory (IRT) is used to quantify the difficulty level of test questions. In order to estimate the difficulty level of questions, test takers (subjects) must answer the questions in advance. However, it is hard to make subjects answer all questions in the item bank. Therefore, a method for estimating the difficulty level of unanswered test questions is proposed that focuses on the question type. This method estimates the difficulty level based on the mix of question types on the test by using IRT to estimate the difficulty level of the unanswered questions. However, the difficulty level of a question may vary by the similarity of answer choices. Therefore, in this paper, a method for estimating the selection probability of each answer choice is proposed that focuses on the similarity of answer choices. Then, a method for estimating the difficulty level of questions by considering the contents of answer choices is discussed.

Important Links:



Go Back