Methods for Imputing Scores When All Responses Are Missing for One or More Polytomous Items: Accuracy and Impact on Psychometric Property

Though a substantial amount of research exists on imputing missing scores in educational assessments, there is little cases where responses or to an item are for all test takers. In this paper, we tackled the problem tests which We considered three missing-data imputation methods—the median method, response theory (IRT) and two-way method—for scores. compared performance these methods with respect their accuracy estimating scaled reliability aforementioned problem. Real data were used comparison. All performed well negligible error: The IRT method provided slightly more accurate most estimates. Recommendations practice provided.