A Comparison of Item Selection Methods Using Response Times in Computerized Adaptive Testing

Abstract

This study aimed to investigate the performance of item selection methods using response times (RT) in terms of the correlations between ability and test speededness and test lengths for computerized adaptive testing with fixed test time. MI, which does not use RT information, seems to estimate the ability parameters more accurately under a lenient time limit and a zero correlation between ability and test speededness condition. However, for MIT, which incorporates RT information in estimation and item selection procedures, the accuracy of ability estimates can be improved compared to MI and MITS as the correlation between ability and test speededness increases and the test length decreases. When item exposure control was applied, the measurement precision would be decreased. Compared to MI, MIT yielded a higher proportion of the number of items for a given test length. With relaxed limits in testing time, the effects of item exposure control on test efficiency would be decreased. Item selection with item exposure control can enhance item pool usage. After applying item exposure control, no significant differences could be expected in the item pool usage.