Effect of person fit index on test psychometric properties

If a person who takes the test answers the difficult item (s) correctly and incorrectly answers the easy item (s), this is unexpected. The degree of this unexpected situation can be determined by person-fit indexes. These indexes, which allow the assessment of individual compliance, have many application areas. In this study, the effect of the test items and the psychometric properties of the person-fit indices on real data set examined. Another aim of the study is to compare the performance of person-fit indexes under different criteria. The real data set was obtained from the achivement scale used in the TIMSS exam. In order to examine the effect of person fit indexes in terms of different criteria, simulative data were produced under many sub-conditions {data compatible with 1PLM and 2PLM; parameter "b" is (-1.5 / + 1.5) and (-3 / + 3) range; mean of parameter "a" is 1, 1.5 and 2; the mismatch rate is 5%, 10% and 20%; 5 different aberrant response type}. One of the findings obtained in the real data is that when around 6% of people who are incompatible with the model are excluded from the test, the test adapts better to unidimensionality. One of the findings obtained in simulative data is that as the "b" parameter range expands, response patterns incompatible with the model create a much more comfortable second dimension, negative factor loadings in the first dimension, and many local dependent item pairs.