Employing differential item function analysis in survey research

Abstract

One of the key assumptions involved in using any survey or measurement is that the instrument works consistently across groups. This is particularly important in survey research where group comparisons are ubiquitous. Differential item functioning analysis examines whether the instrument systematically biases in favor of one group. The findings from such an analysis are unattainable in traditional approaches to examining instrument validity, and yet, it is rare to find DIF analysis in surveys. This process illustrates DIF analysis with logistic regression using the Faculty Survey of Student Engagement. We find FSSE items did not show the presence of DIF. This provides confidence to users of this instrument that it measures the same constructs in the same way across different groups.