Zero-inflated poisson regression untuk memodelkan faktor-faktor yang mempengaruhi terjadinya kebakaran di kabupaten Sidoarjo

Abstract

The purpose of this research is determine the factors that affect the rate of fires in Sidoarjo using analysis Zero-Inflated Poisson (ZIP) regression. The procedure in this research through the steps are detecting the distribution of a variable number of fire (Y), identified cases multikolinearitas, establishing models Poison regression, testing overdispersi or underdispersi, establishing models ZIP regression, testing coefficient simultaneously and partially, and choose a suitable model. The independent variable used are population density (X_1), the number of factories (X_2), sugarcane and paddy land area (X_3), vacant land that overgrown with weeds (X_4), and year (X_5) Based on the regression model Zero-Inflated Poisson (ZIP) obtained best model of rate of fires that affected by the population density (X_1), sugarcane and paddy land area (X_3), and the year (X_5) to distinguish the difference the incidence of fires in every year. The model has a value of AIC (257,4) and the smallest value of R^2 (70,86).