Abstract
This paper describes a new and efficient method of respondent's ability parameter estimation using answer-until-correct (AUC) procedures in computer-based testing. The AUC procedure for multiple choice items allows second and subsequent choices from the distractors until the correct choice is made, if the subject's first choice is incorrect. The model used for this estimation is an extension of the nominal response (NR) model of item response theory (IRT). The characteristic and information functions of the AUC response paths are formulated in this paper. Simulated data were used to investigate the efficiency of the AUC model. Data were generated for 1, 300 respondents with tests having 9, 11, 13, 17 or 21 items with three alternative choices. The ability parameters estimated by the AUC model were more accurate than those estimated by the NR model. These results confirm that the AUC model can estimate ability parameters as accurately as the NR model and requires fewer test items.
