Beyond Average Sex Differences: A Bayesian Analysis of Extreme and Discordant Physical Activity–Dietary Behavioral Profiles in University Students

Background/Objectives: Physical activity (PA) and dietary behaviors are commonly examined separately or through average associations, potentially obscuring individual heterogeneity and behavioral discordance. This study aimed to identify extreme and recurring PA–dietary behavioral profiles and examine sex-related differences. Methods: This cross-sectional study included 418 university students (199 men and 219 women). Physical activity was assessed using the International Physical Activity Questionnaire (IPAQ-LF), whereas dietary behaviors were evaluated using the Questionnaire of Eating Behaviors (QEB) and expressed as the pro-Healthy Diet Index (pHDI-8) and non-Healthy Diet Index (nHDI-8). Robust Mahalanobis distance (RMD) was used to identify extreme multivariate configurations. Bayesian finite-mixture modelling was subsequently used to explore recurring behavioral profiles, and concordant and discordant PA–dietary configurations were examined by sex. Results: RMD identified 49 participants (11.7%) with extreme multivariate configurations; these configurations were significantly more frequent among men than women (15.1% vs. 8.7%, p = 0.042). Bayesian modelling yielded a converged, exploratory two-profile solution differentiated primarily by dietary behaviors. The more favorable dietary profile included 37.1% of participants and was more frequent among women (54.8% vs. 17.6%; OR = 5.68, 95% CI: 3.62–8.92). Discordant PA–dietary configurations occurred in 49.5% of the sample and were more frequent among women (57.1% vs. 41.2%; OR = 1.90, 95% CI: 1.29–2.80). Women predominantly exhibited lower PA and more favorable dietary behaviors, whereas men more frequently exhibited higher PA and less favorable dietary behaviors. Discordance was not associated with multivariate extremeness (p = 0.823). Conclusions: Joint assessment of PA and dietary behaviors revealed heterogeneity beyond average sex differences. Discordant PA–dietary configurations differed by sex, although their distribution was sensitive to reference cutoffs. Discordance and multivariate extremeness represented distinct aspects of behavioral heterogeneity. These findings support person-centered assessment of lifestyle behaviors and may inform more targeted health-promotion strategies.