Developments in areal interpolation methods and GIS
- Cite this article as:
- Flowerdew, R. & Green, M. Ann Reg Sci (1992) 26: 67. doi:10.1007/BF01581481
- 472 Downloads
This paper is a review and extension of the authors' research project on areal interpolation. It is concerned with problems arising when a region is divided into different sets of zones for different purposes, and data available for one set of zones (source zones) are needed for a different set (target zones). Standard approaches are based on the assumption that source zone data are evenly distributed within each zone, but our approach allows additional information about the target zones to be taken into account so that more accurate target zone estimates can be derived. The method used is based on the EM algorithm. Most of the work reported so far (e.g. Flowerdew and Green 1989) has been concerned with count data whose distribution can be modelled using a Poisson assumption. Such data are frequently encountered in censuses and surveys. Other types of data are more appropriately regarded as having continuous distributions. This paper is primarily concerned with areal interpolation of normally distributed data. A method is developed suitable for such data and is applied to house price data for Preston, Lancashire, starting with mean house prices in 1990 for local government wards and estimating mean house prices for postcode sectors.