Assessing the accuracy of kernel smoothing population surface models for Northern Ireland using geographically weighted regression

Abstract

This paper makes use of two sets of Census outputs for Northern Ireland – counts by (i) output areas (OAs) and (ii) 100 m grid cells. Population surfaces are derived from OAs using a kernel smoothing approach and the accuracy of these surfaces is assessed using the 100-m grid square product. The key contribution of this paper is in pulling together the benefits of a gridded population data source with local regression procedures to provide a more detailed assessment of surface modelling accuracy than possible in any previous studies.

Keywords:

• Population surface modelling
• spatial variation
• census
• ordinary least squares (OLS)
• geographically weighted regression (GWR)

Acknowledgements

Provision of the grid square data by the Census Office (part of the Northern Ireland Statistics and Research Agency – NISRA) is acknowledged.

Notes

1. https://www2.geog.soton.ac.uk/users/martind/davehome/software.htm.