ESTIMATION PARAMETER GEOGRAPHICALLY WEIGHTED ZERO INFLATED POISSON REGRESSION (GWZIPR) WITH FIXED BISQUARE KERNEL

Abstract

Poisson regression is a form of regression analysis used to model data in the form of count (number). Poisson regression method requires the existence of equidispersi that is the condition where the mean and variance of the response variable is equal. But sometimes there is an overdispersion phenomenon in the data modeled with the poisson distribution. Overdispersion means that data has a variance greater than the mean. Overdispersion shows that there is a population heterogeneity. Consequently, the estimation of the parameters on the data under such conditions becomes imprecise. One method to overcome the overdispersion is the Zero Inflated Poisson regression. Then the development of the ZIP regression that has taken into account the spatial factor is called Geographically Weighted Zero Inflated Poisson Regression (GWZIPR). The parameter estimation of the GWZIPR model is carried out by the Maximum Likelihood Estimation (MLE) method and completed using the Expectation-Maximization (EM) algorithm. Weighting function used is fixed bisquare kernel. This study describes the factors that influence the disease Tetanus Neonatorum in all districts / cities in East Java Province.