Comparative Modeling of Hand Grip Strength in Malaysian Young Adults: From Classical to Allometric Regression to Machine Learning Approaches

This comparative study investigates hand grip strength (HGS) in young Malaysian adults, utilising classical, allometric, and machine learning (ML) regression techniques to identify key predictors and establish population-specific data. We conducted a comparative analysis of HGS and anthropometric variables using four distinct regression methods: stepwise regression, allometric regression, LASSO regression, and Random Forest, allowing for a comprehensive assessment of predictive power and the identification of optimal scaling relationships. The findings consistently identified Forearm Circumference as the most significant predictor of HGS across all models, with Palm Circumference and Length of Palm-Wrist also being key determinants. While traditional linear models provided statistically significant results, the Random Forest models demonstrated superior predictive accuracy, with R-squared values ranging from 0.44 to 0.49, supporting the utility of ML in capturing complex, non-linear relationships in biomedical data. Ultimately, this research establishes a foundational understanding of HGS determinants in a previously under-researched demographic, providing valuable normative data for young Malaysian adults with implications for fields such as ergonomics and public health.