MODELING LONGITUDINAL FLOOD DATA IN WEST SUMATRA USING THE GENERALIZED ESTIMATING EQUATION (GEE) APPROACH

Abstrak

Flooding is one of the many natural disasters that often hit Indonesia. In July 2023, three areas in West Sumatra experienced floods and landslides which caused damages and even 2 missing victims. Since November 16th, 2023, 8 hamlets in Meranti Village, Landak District, West Sumatra have been inundated by floods which affected families and many public facilities. This research uses data from West Sumatra Province Central Statistics Agency. The data used is 2014, 2018 and 2021. The response variable used is the number of villages/sub-districts experiencing natural disasters according to district/city ( ). The predictor variables used are regional topography, the number of water channels such as rivers, reservoirs, etc., the number of fields cleared through burning, the number of villages/sub-districts in C excavation area, and the number of dumpsters. This research uses Negative Binomial Regression with the Generalized Estimating Equation (GEE) approach. In the Poisson regression test, the QIC value based on Independent Working Correlation Structure (WCS) is with deviance value of, degree of freedom of, and dispersion score of 4,6144. Because the dispersion value is greater than 1, it can be concluded that there is overdispersion. Because there is more than one overdispersion, it is overcome by using negative binomial. The results of parameter estimation using negative binomial regression based on Independent WCS showed that only one variable was significant, which is the number of fields cleared through burning with deviance value of, degrees of freedom of and a QIC of. Negative Binomial regression model that was formed is ). From the two regression models used, namely Poisson and negative binomial, it was found that the negative binomial regression model was the best model because it had the lowest QIC value of.