B0411
Title: Efficient corrections for standardized person-fit statistics
Authors: Kylie Gorney - Michigan State University (United States) [presenting]
Sandip Sinharay - Educational Testing Service (United States)
Carol Eckerly - Educational Testing Service (United States)
Abstract: In educational and psychological testing, person-fit statistics are used to identify individuals who are displaying aberrant, or unusual, behaviour. Many popular person-fit statistics belong to the class of standardized person-fit statistics, $T$, and are assumed to have a standard normal null distribution. However, in practice, this assumption is incorrect since $T$ is computed using (a) an estimated ability parameter and (b) a finite number of items. Several corrections have been suggested to improve the accuracy of person-fit statistics. However, all of these corrections are limited in that they are either computationally intensive (i.e., they require the simulation and analysis of several large data sets) or they account for (a) or (b), but not both. Three new corrections are proposed that are computationally efficient and account for both (a) and (b). Detailed simulations reveal that the new corrections outperform existing corrections by being able to control the Type I error rate while also maintaining reasonable levels of power.
