Spatial probit regression model: Recursive importance sampling approach

The probit regression model is a model used to analyze the relationship between categorical response variables, with predictive variables that are numerical, categorical, or the combination of both. In some cases, the data which generated in response variables in the categorical probit regression may also be influenced by spatial autocorrelation. If the influence of spatial autocorrelation is ignored and still using the standard probit model, it will produces an inconsistent and biased parameter estimation. To handle the case, it has developed a method of probit analysis that has considering the spatial element in it is spatial probit regresion. According to several estimation methods that using spatial probit regression, the recursive importance sampling (RIS) method performs best or surpass other methods in terms of accuracy. The formation of spatial probability regression model that using RIS method was applied on the case of health index in Papua Island, Indonesia, where significant predictor variables in establishing spatial probability regression model on health index were; per capita expenditure, percentage of households with proper drinking water source, percentage of population having smoking habits, average length of school, and ratio of health facilities to the number of villages. The model has an accuracy value of 61.90 percent.