Clusterwise Regression Model for Statistical Downscaling to predict Daily Rainfall using Gamma Distribution

Abstract

This study presents the development of clusterwise regression with gamma distribution and normal distribution. This clusterwise regression model is expected to be a solution for high variance data in order to minimize the error rate in modeling. The first study is simulation of clusterwise regression with gamma distribution and normal-gamma mixture. This simulation is divided into two-cluster and three-cluster simulations with gamma distribution, normal distribution, and mixed normal-gamma distribution with different parameters. The application of clusterwise regression was carried out at four rain stations in West Java, including Bandung, Bogor, Citeko, and Jatiwangi rain stations by first doing variable reduction using principal component analysis.