Covid-19 Modeling towards socioeconomic and health data from New South Wales (NSW) -- Australia: An approach via Geospatial Analysis and Geographically Weighted Poisson Regression (GWPR)

TL;DR

This study employs geospatial analysis and Geographically Weighted Poisson Regression to model COVID-19 relationships with socioeconomic and health factors across local government areas in New South Wales, revealing nonstationary relationships and the importance of local-scale modeling.

Contribution

It introduces a combined geospatial and GWPR approach to analyze COVID-19 data with socioeconomic and health variables at a local level in NSW, Australia.

Findings

Positive relationships between COVID-19 and population, cancers, and age group 60-85.

Negative relationship observed with ischaemic heart disease, but coefficients are near zero.

GWPR model shows R2 between 45-73%, indicating good fit and nonstationary relationships.

Abstract

An integrated approach of spatial data analysis and Geographically Weighted Poisson Regression (GWPR) along with global regression techniques are used in this study. This approach aims to model relationships between dependent variable Covid-19 and independent variables from socioeconomic and pre-existing health conditions within the local government area (LGA) in New South Wales (NSW)-Australia. Based on geospatial data analysis and a step-by-step procedure in building both global and GWPR models, four (4) independent variables are finally selected to investigate relationships between dependent and independent variables at the local scale. The GWPR model's results with the Goodness-of-Fit (R2) range between 45-73% exhibit positive relationships between Covid-19 and the total population, the cancers, and the people with ages between 60 and 85 in most of the NSW state. Meanwhile, a negative relationship is observed between Covid-19 and the ischaemic heart disease; however, the estimated coefficients for this relationship are very low and close to zero; hence further investigation, including assessment from a different perspective, is necessary for validation. In conclusion, the model suggests that the relationships between the dependent variable and independent variables are nonstationary. Therefore, GWPR model calibration plays a vital role in geographic modelling at the local scale.