A practical framework for selecting and interpreting regression models for binary outcomes in social science and public health studies

In social science and public health research, binary outcomes are common (e.g., disease presence, mental health status, health behaviors), as they simplify complex constructs into clear categories, facilitating interpretation, communication, and application in clinical or policy decision-making. Yet methodological missteps persist, particularly the misuse of logistic regression when binary outcomes are common and the use of unclear statistical language. This commentary discusses four frequent modeling and terminological errors, illustrates how they hinder valid inference in the social science and public health discipline, and proposes a practical framework for selecting and reporting models in cross-sectional studies of social and health phenomena.