Model Regresi Nonparametrik Spline Truncated Untuk Memprediksi Prevalensi Stunting di Sumatera Barat

Abstract

Not all predictor variables can be approached with a parametric regression approach. This occurs because there is no information about the form of the relationship pattern between the response variable (Y) and the predictor variable (X). Therefore, a nonparamatric regression approach must be used to estimate the regression curve. Truncated Spline nonparametric regression has high flexibility and is able to handle changing data behavior in certain sub-intervals and tends to find its own data estimate wherever the data pattern moves. This study aims to model the prevalence of stunting in West Sumatra using truncated spline nonparametric regression. Model selection is carried out with 1 knot, 2 knots and 3 knots. The research results show that the best model is using 3 knot points, with a resulting GCV of 0.01 and a coefficient of determination (R2) of 99.99%.