A modified test in flexibly shaped spatial scan statistic for detecting poverty hotspots.

Abstract

A flexibly spatial scan statistic was proposed by Tango and Takahashi (2005) that can detect arbitrary shaped clusters (hotspots). In this research, a flexibly shaped spatial scan statistic will be applied to perform poverty hotspots detection in Bogor. Overall population of family in Bogor is 1,094,480 families and 29.44% of families who live in Bogor are categorized as poor family. The problem was flexibly spatial scan statistic use Poisson distribution in their calculation but the poverty rate was too high to be approached by Poisson distribution. In fact that Poisson must be constructed from the number of trials is large and the probability of success is small. We proposed Bernoulli distribution instead of Poisson. This distribution was more suitable to the faced case. That have to be noticed is sampling unit from both the distribution. At Poisson, what becoming sampling unit is village while at Bernoulli the sampling unit is family. Each family has only two possibility status, that are poor and not poor. It was meets the characteristic of Bernoulli distribution. In this proposed method, the events of being poor are independents distributed, so we could find the likelihood function as multiplication of each individual mass function. In this research, the proposed test tends to be more sensitive compared to Septiani (2006), we obtained more significant hotspot. The order of log likelihood ratio statistic was slightly different, there were only minor changes. Septiani’s result using Poisson distribution still relevant to detect poverty hotspots in Bogor.