The effect of missing data on the bias and accuracy of academic alienation questionnaire questions

Abstract

One of the important type of errors is differential item functioning (DIF). The aim of this study, using the Monte Carlo simulation method, was to determine whether the effect of missing data on the bias and precision of the Likelihood Ratio test for DIF detection differ across different missing data methods? The factorial design was used to examine the effect of six MDM methods on the effectiveness of IRT-LR test for DIF detection in the GRM, in terms of Type I error and statistical power. The statistical population in this study was all university students with undergraduate degrees in educational sciences and psychology in Tehran. A sample of 1100 students from this community were among the students studying at Allameh Tabataba’i, Tehran, Shahid Beheshti, Kharazmi, Islamic Azad University Tehran North Branch, Islamic Azad University Tehran South Branch and Islamic Azad University Central Tehran Branch in 2019. The academic alienation questionnaire was administered to them. G2 index and the effect size η2 were calculated for all first-order interactions. Multiple imputation and single regression substitution methods had the lowest contribution in the control of the type I error and the methods based on maximum likelihood and the person mean substitution showed the highest contribution in terms of error control, respectively. The statistical power of the test was similar in all methods in this study. The results of this study indicated that the missing data and observations and the length of the test had an important role in the effectiveness (bias and precision) of the test.

Keywords

• : missing data
• academic alienation
• likelihood ratio test
• type i error rate
• statistical power