Abstract:

Person-fit statistics (PFS) refers to the consistency of either an individual’s response pattern within a given measurement model or other examinees’ response patterns within a given sample. If we can distinguish the aberrant item score patterns, the test score should be more interpretable, and, consequently, the reliability and validity should be heightened. However, nowadays most of the researches in person-fit statistics have been concentrating on ability tests and computer adaptive tests instead of personality questionnaires. The purpose of this research is to investigate the effect of person-fit statistics in the reliability, inner structure validity, and the criterion-related validity of personality questionnaire. Eysenck Personality Questionnaire (EPQ) was used in the present study. Firstly, we removed L scale and got the positively and negatively worded EPQ questionnaire for the same group of participants. Then, we merged all the items and conducted the exploratory factor analysis, and removed the items whose load was less than 0.3. Finally, we selected E and N subscales for the study. We chose the number of items, where response vector in positively worded items was unequal to that in the negatively worded items (called unequal index Y), as the criterion of estimating the person-fit statistics. Eight person-fit statistics (G, C, MCI, U3, U, W, ECI6, and L) were chosen to investigate influences to the reliability and validity of the tests, and to analyze the correlation between the person-fit statistics and unequal index. Then the effectiveness of PFS in personality questionnaire was investigated and the power of the eight person-fit statistics was compared. As shown in the results, the reliability was increased after deleting 5%, 10%, and 20% aberrant individuals. The reliability of the half of the good-fit participants was more significant than that of the bad-fit participants. L, C, and G were more sensitive to reliability than other PFS. The results were similar for E scale and N scale. Although the criterion-related validity didn’t steadily increase after deleting 5%, 10%, and 20% aberrant individuals, the criterion-related validity of good-fit participants were more significant than that of the bad-fit participants. All of the PFS showed significantly effect to the inner structure validity. L displayed an obvious and stable effect to the inner structure validity. The eight PFS had different correlation levels with Y, and the correlation coefficient among L, G, and Y was the highest one. Therefore, these results suggest that PFS can rectify the reliability and validity. In other words, PFS is suitable to personality tests and L based on IRT is the most ideal one.

Key words: Person-Fit Statistics, Aberrant item score pattern, Personality Test, EPQ

摘要：

个人拟合指标是考察心理测验中偏差得分模式的新方法。研究中考察了G、C、MCI、U3、U、W、ECI6、L等8个拟合指标对艾森克人格问卷信效度的影响，以及各指标与正反向题回答不一致项目数的相关。结果表明，删除不同比例拟合程度不好的个体后，测验的信效度明显提高。同时PFS可鉴别人格测验中的默认反应偏差。各指标中l对测验信效度的改善效果最为理想。

关键词: 个人拟合指标, 异常得分模式, 人格测验, EPQ