Detection of Differential Item Functioning Magnitude in Psychological Measurement with Missing Data

Abstract

The paper investigated the effectiveness of missing data methods in detecting differential item functioning magnitude in polytomous scored non-cognitive items with a view to determining different levels of magnitude existing in non-cognitive items and the difference in the ability of missing data methods to detect DIF magnitude. Using the sample of 1,500 senior secondary school students, drawn through multistage sampling technique from Osun State, data were collected with the Achievement Motivation Inventory (AMI). The result showed that with Full Information Maximum Likelihood (FIML) 81.3% possess small DIF magnitude, while 4 (12.5%) items possess moderate DIF magnitude, while high DIF magnitude occurs in 2(6.3%) items, and 24 (75.0%) were categorised as having small DIF magnitude, 2 (6.3%) with moderate DIF magnitude, while 6 (18.8%) was classified as having high magnitude of DIF using Multiple Imputation (MI). The result further revealed that item effect size (magnitude) across different methods did not distinctly differ from one another (X2 = 0.16, df = 1, p > 0.05). The study concluded that the two missing data methods were effective in detecting magnitudes of DIF present in polytomous scored non-cognitive items, and the differences that exist in the abilities of these missing data methods is not statistically significant.