Date of Award

Summer 2021

Document Type

Open Access Thesis

Department

Statistics

First Advisor

Stella Self

Abstract

Spatial clustering detection methods are widely used in many fields of research including sociology, epidemiology, ecology, and criminology. The objective of this study is to assess the performance of four spatial clustering detection methods: the average nearest neighbor ratio, Ripley’s K function, local Moran’s I and Getis-Ord Gi* statistics. We conduct a simulation study to evaluate the performance of each method for areal data under different types of spatial dependence and three different areal structures; a 20x20 regular grid, United States counties in six states and Canadian forward sortation areas (FSAs) in three provinces. The results shows that the empirical type I error rates are inflated for ANN and Ripley’s K. For local Moran’s I and Getis- Ord Gi* statistics empirical type I error rates are less than or equal to 0.05 for most of the units in all three areal structures and classification accuracy is closer to 1. We find that the performance of ANN and Ripley’s K are not reliable when applied to areal data unlike local Moran’s I and Getis-Ord Gi*.

Rights

© 2021, Nadeesha Dilhani Vidanapathirana

Included in

Biostatistics Commons

Share

COinS