What do we really mean by gender differences? The statistical construction of difference in educational research
Grade differences between males and females are frequently reported in educational research as though they constitute a straightforward property of students; one group performs better than another. Yet what counts as a “difference” depends in part on how the outcome is represented statistically. This paper examines this issue using student grades from an introductory statistics course as an empirical illustration. The same data are analyzed using several common approaches, including a t -test, chi-square test, Mann–Whitney test, and cumulative link model. Although the analyses converge on a broadly similar empirical pattern, they represent the difference in substantially different ways, as a difference in means, distributions, rank, or odds. The example illustrates that statistical models do not merely quantify differences that exist independently of analysis; they also define which features of an outcome are treated as relevant to the difference. We argue that greater precision about these choices can help educational research distinguish more clearly between what statistical analyses establish and what researchers subsequently infer from them.
Authors
- Marius Ole Johansen (ORCID: https://orcid.org/0000-0002-4289-9048)
- Kristoffer Skomsøy Fjællingsdal (ORCID: https://orcid.org/0000-0001-5370-6937)
Institutions
- Norwegian University of Science and Technology (NO)
- University of Bergen (NO)
Publication Details
- Journal
- Frontiers in Education
- Published
- 2026-11-17
- DOI
- https://doi.org/10.3389/feduc.2026.2001702
- Primary Topic
- Statistics Education and Methodologies
- Type
- article
- Field-Weighted Citation Impact
- 0.00