Abstract
An comparison of the Area Methods, Lord's chi-square test and Logistic Regression Analysis for Assesing Differential Item Functioning. The present study compared the performance under different conditions of the signed area measure Z(AES), the unsigned area measure Z(AEA), Lord's chi-square test and the logistic regression analysis (LR) on detection of differential item functioning (DIF). Sample size, amount and type DIF, difference in the group trait level averages, percent of DIF items test and significance levels were manipulated. The results show that Lord's test and Z(AEA) were effective in assessing uniform, non-uniform and mixed DIF. The LR procedure was low power for detecting uniform DIF, however it was able to detect non-uniform and mixed DIF. When the sample size, amount of DIF and percent of DIF items were larger high number of false positives were obtained using the Lord's statistic, Z(AES) and Z(AEA).
