Abstract

This paper investigates the presence of Differential Item Functioning (DIF) in ordinal data. Motivated by a dataset of 1007 respondents collected at the Laboratory for Applied Statistics (LASt), University of Naples Federico II, we analyze gender differences in a set of Likert-scale items measuring attitudes toward gender roles and responsibilities, scientific aptitude, and workplace dynamics. We develop a methodological framework for the detection, decomposition, and interpretation of DIF in ordinal models, combining global Wald tests, parameter-specific tests, and comparative examination of item parameters. The findings reveal substantial threshold DIF for several items associated with traditional gender-role expectations, indicating differential use of the response scale across males and females even at equal latent trait levels. Implications for measurement invariance, interpretability, and survey design in social science applications are discussed.