Detecting Spatial Autocorrelation for a Small Number of Areas: A practical example

Abstract

Moran's I is commonly used to detect spatial autocorrelation in spatial data. However, Moran's I may lead to underestimating spatial dependence when used for a small number of areas. This led to the development of Modified Moran's I, which is designed to work when there are few areas. In this paper, both methods will be presented. Many R programs enable calculating Moran's I, but to date, none have been available for calculating Modified Moran's I. This paper aims to present both methods and provide the R code for calculating Modified Moran's I, with an application to a case study of dengue fever across 14 regions in Makassar, Indonesia.