Title: Modeling Toddler Weight Growth in Banyuwangi Based on Z-Score Curves Using Local Linear Regression for Health Assessment
Authors: Nur Chamidah, Ardi Kurniawan, Toha Saifudin, Naufal Ramadhan Al Akhwal Siregar, Vanisia Suci Ananda, Ratna Anugrahaningtyas, Amelia Fatihah, Nararya Raissa Tiara Putri
Volume: 10
Issue: 3
Pages: 9-16
Publication Date: 2026/03/28
Abstract:
Children represent an important generation that will influence the direction of national development; therefore, their growth must be monitored carefully. Stunting remains a significant public health concern in Indonesia, including Banyuwangi Regency. Although the prevalence decreased from 8.64% in 2021 to 2.44% in 2024, more than 2,200 toddlers are still projected to experience stunting by 2025. This study models toddler weight growth in Banyuwangi Regency using local linear nonparametric regression based on Z-scores. The analysis is based on primary data from 1,627 toddlers aged 0-60 months across nine villages. The estimation is performed using the Gaussian Kernel (GaussK) method with bandwidth selected through Cross-Validation (CV). The results show low prediction errors across all Z-score levels. For male toddlers, MAPE values range from 1.28% to 3.71%, with Rē between 0.983 and 0.996. For female toddlers, MAPE ranges from 1.51% to 2.50%, with Rē between 0.976 and 0.995. Based on the MAPE criterion (? 10%), the estimates fall within the highly accurate prediction category. The smallest errors are observed around the median Z-score (0 SD), while relatively higher variability appears at extreme Z-score levels. These findings indicate that the Gaussian Kernel local linear regression with CV bandwidth selection provides consistent and reliable estimates of toddler weight growth. The results may serve as a statistical reference for evaluating growth patterns and supporting evidence-based nutritional interventions in Banyuwangi Regency.