The Future of Education

This study aimed to identify latent profiles of relational risk in school trajectories and to examine whether these profiles represent empirically distinct groups, combining person-centered and group-comparison approaches.

The sample comprised 3536 public school students enrolled from the 9th grade to the 3rd year of upper secondary education in four Brazilian states (Rondônia, Minas Gerais, Mato Grosso, and Maranhão). Participants completed the IAFREE-A, a multidimensional instrument designed to assess protective and risk-related factors in school trajectories across student–student, student–school, student-school professionals, student–family, and student–community domains. Latent Profile Analysis (LPA) was conducted using scores from the Student-Student (SSt) dimension and its three factors (interpersonal relations and social skills, educational expectations, and belonging/identification). Model selection followed an analytic hierarchy process based on multiple fit indices (AIC, AWE, BIC, CAIC, CLC, and KIC). The best-fitting solution corresponded to Model 2 with three latent classes, presenting high entropy (0.84), indicating good classification accuracy and clear separation between profiles. The profiles were interpreted as low, medium, and high relational risk.

To test whether the identified profiles were substantively distinct, non-parametric Kruskal–Wallis analyses were conducted comparing the three classes across the SSt dimensional score, its factors, and other dimensions of the instrument. Results showed statistically significant differences between all classes for the SSt total score (χ² = 2912, p < .001, ε² = 0.824), as well as for interpersonal relations and social skills (ε² = 0.651) and belonging/identification (ε² = 0.639), indicating very large effects. Differences in educational expectations were also significant, with a moderate effect size (ε² = 0.158). Additionally, the profiles differed significantly across the SSc dimension, with large effect sizes (ε² = 0.350), especially for the factor Perception of the School as a Safe Place, as well as for institutional responses to discrimination, suggesting that higher relational risk among students is closely linked to perceiving the school as unsafe and insufficiently protective. All post hoc comparisons confirmed significant differences between low, medium, and high risk groups.

Together, these findings provide convergent evidence that the latent profiles reflect qualitatively distinct configurations of relational risk. Peer relationships and school belonging emerge as central axes through which vulnerabilities accumulate and extend to broader institutional experiences, supporting a process-oriented understanding of school trajectory risk.

Keywords: latent profile analysis; school trajectories; peer relations; educational risk; Kruskal–Wallis.

References:
R Core Team (2024). R: A Language and Environment for Statistical Computing. https://cran.r-project.org
Rosenberg, J. M., Beymer, P. N., Anderson, D. J., Van Lissa, C. J., & Schmidt, J. A. (2021). tidyLPA [R package]. https://CRAN.R-project.org/package=tidyLPA
Seo, H. J. (2025). snowRMM [jamovi module]. https://github.com/hyunsooseol/snowRMM