The Effectiveness of (Lz*new) Index in Detecting Person Aberrant Response Patterns According to IRT

Abstract

The present study aimed to investigate the effectiveness of Lz*new index in detecting person aberrant response patterns, according to the Item Response Theory (IRT), by investigating the empirical type I error in the light of the different logistic models (2PLM and 3PLM) of IRT at 11 different ability levels, and three different levels of statistical significance. To achieve the objectives of the study, binary data generated from a 40-item test were used. Up-to-date statistical software packages (Wingen2, Bilog- MG3, R-Fit, and SPSS) were used chronologically to analyze the data. Findings of the study indicated the effectiveness of Lz*new index in detecting person aberrant response patterns, no matter what the type of the logistic model used. However, it was more effective in the 3–Parameters Logistic Model (3PLM) than the 2–Parameters Logistic Model (2PLM). The results also indicated that type I error (3PLM) rates close to the nominal value for most ability level, no matter what the level of statistical significance was. In the light of the study findings, a number of conclusions and recommendations are suggested.