Geographically weighted local statistics applied to binary data

Abstract

This paper considers the application of geographically weighting to summary statistics for binary data. We argue that geographical smoothing techniques that are applied to descriptive statistics for ratio and interval scale data may also be applied to descriptive statistics for binary categorical data. Here we outline how this may be done, focussing attention on the odds ratio statistic used for summarising the linkage between a pair of binary variables. An example of this is applied to data relating to house sales, based on over 30,000 houses in the United Kingdom. The method is used to demonstrate that time trends in the building of detached houses vary throughout the country.