Mixed geographically weighted regression model (case study: The percentage of poor households in Mojokerto 2008

Regression analysis is a statistical analysis that aims to model the relationship between response variables with predictor variables. Geographically Weighted Regression (GWR) is statistical methods used for analyzed the spatial data in local form of regression. Where certain predictor variables influencing the response are global while others are local used the Mixed Geographically Weighted Regression (MGWR) model to solve the problem. The results showed that Weighted Least Square (WLS) can be used to estimate the parameter model and Cross Validation (CV) for the selection of the optimum bandwidth. Goodness of fits tests for a global regression model and MGWR approximated by F distribution as well as on the test of global parameters and local parameters simultaneously and for testing the partial model parameters using the t distribution. The applications of MGWR model in the percentage of poor households in Mojokerto showed that MGWR model differs significantly from the global regression model. Based on Akaike Information Criterion (AIC) values between the global regression model, GWR and MGWR model, it is known that the MGWR model with a weighting Gaussian kernel function is the best model used to analyze the percentage of poor households in Mojokerto (2008) because it has the smallest AIC value.