A Proposal of Uniform Adaptive Testing for Reducing Measurement Error Bias

DOI IR

Bibliographic Information

Other Title
  • 測定精度の偏り軽減のための等質適応型テストの提案

Abstract

Adaptive testing is a question format of computer testing that estimates an examinee's ability sequentially and which produces question items with the highest estimation accuracy according to the examinee's ability. The technique mitigates the creation of overly easy or overly difficult questions, which can reduce the time spent on a test, and reduces the number of items without reducing the estimation accuracy for the examinee's ability. However, in conventional adaptive tests, it is highly likely that the exact same group of items will be prepared for examinees who have equivalent ability. The tests cannot be used practically under circumstances by which the same learner can take a test multiple times, such as SPI and GTEC. In this paper, we propose a multiple equivalent adaptive test that adaptively creates different items for examinees even if those with equivalent capabilities, maintaining the same evaluation accuracy. Specifically, we follow the procedure outlined below. 1) We compose an item cluster for a multiple equivalent test based on the amount of test information so that the measurement accuracy for examinees'ability can be equivalent despite consisting of different items. To compose a multiple equivalent test, we use a technique that employs the maximum clique problem to maximize the number of compositions from items within an item bank. 2) Regarding an item cluster for a multiple equivalent test as an item bank, we propose a multiple equivalent adaptive test that estimates the value of an examinee's ability sequentially and which selects items with the greatest amount of information for the value of ability from an item cluster for a multiple equivalent test. This paper presents the effectiveness of the technique through a simulation experiment and with item banks used by actual test providers.

Journal

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top