Some new methods to detect person fit in CAT

Person fit is concerned with detecting nonfitting item-score patterns. Most person-fit statistics have been proposed in the context of conventionally administered tests or paper-and-pencil (P&P) tests. In this study, we will first review some existing person-fit studies in a computerized adaptive testing (CAT) context and then investigate the usefulness of some new fit statistics that are based on the specific characteristics of a CAT. Both the use of statistical process control and the use of nonparametric tests is explored. The results of a simulation study to detect nonfitting response patterns in a CAT showed that the detection rate of these statistics is comparable to the detection rate of person-fit statistics in P&P tests.