Parameter Estimation of Spatial Error Model – Multivariate Adaptive Generalized Poisson Regression Spline

The non-parametric regression method becomes an alternative that prioritizes flexibility. Therefore, it is possible to obtain a regression curve model when its shape is not yet known. Multivariate adaptive regression spline (MARS) is one of the non-parametric approaches. In 1991, MARS was introduced by Friedman. The MARS approach, which uses nonparametric regression, can consider additive and interactive effects between predictor variables. MARS modeling has typically been used to model continuous or categorical data. However, researchers in the health sector not only encounter data with continuous or categorical responses but also count data. The original MARS method did not support count data with varying variances and means. Therefore, this study aims to develop the Spatial Error Model-Multivariate Adaptive Generalized Poisson Regression Spline (SEM-MAGPRS), which combines the MARS method with the generalized Poisson regression method with spatial effects.