Abstract
The rising prevalence of mental health difficulties among higher education students is a widely documented concern, with financial stress consistently identified as a contributing factor 1. However, outliers in observational data can substantially distort the estimated relationship between financial strain and mental health outcomes when conventional ordinary least squares (OLS) regression is used 2. This study presents a comparative robust regression analysis of the relationship between financial stress and a composite mental health index among university students, contrasting OLS with M-estimation (bisquare and Huber weighting), S-estimation, and MMestimation, alongside two outlier-treatment strategies (case deletion and mean imputation). Analysis of 2,549 student records shows that the OLS estimate is severely attenuated by outliers (R² = 0.042), whereas the robust methods yield a substantially stronger association (R² ≈ 0.17-0.20) with slopes 23-44% steeper than the OLS estimate. M-estimation with bisquare weighting and MM-estimation produced almost identical coefficients, indicating stable estimation 3. These findings indicate that unaddressed outliers can mask a genuine association between financial stress and student mental health, underscoring the value of robust regression for accurate inference in this area of mental health research.