The Power Of Person-fit Statistics To Detect Different Bias Behavior In Personality Test

How to detect the bias response in types of psychological test has attracted many researchers'attention for a long time. Recently, researchers use Person-Fit Statistics to detect the bias response in Personality Test, which was regarded as an important psychological measurement index. Person-fit statistics (PFS) refer to the consistency of an individual's response pattern, ether with a given measurement model or with other examinees'response patterns within a given sample. Although there were many kinds of PFS, we have to detect which one is better in detecting bias response.Four IRT-based appropriateness indices for detecting aberrant response patterns were employed in this study, they were ECI4,LnU, LnW and Lz, which had performed well in previous researches. The item-response data were generated under the two-dimensional model used Monte Carlo procedure in SAS 11.0, and discussed the detection power under the types of bias, proportion of deviation, and test length. The research had two parts and compared the detection power of four indices under the true ability and estimate ability circumstance.Results showed that index LnU and LnW can detect desirable response and faking bad responses well, but indices of ECI4, LnU and LnW showed little power when detecting the random response; Index Lz can detect all the three kinds of bias, and the detection power of random response become higher in long test. The detection ratio was reduced by the increasing of bias percentage. But the increasing of the items enhances the detection power of all indices, especially indices ECI4 and Lz. Totally, PFS can detect bias response in Personality test, the relative effectiveness of theses person-fit indices depended on test length, the proportion of deviation items and types of bias. Index Lz is the most fitful to detect bias behavior in Personality Test.