Spatial analysis and modelling deals with analysis, modelling and estimation procedures that take account of geographical (spatial) relationships.
The gains from spatial analysis for the Office for National Statistics (ONS) potentially include better local estimates for resource allocation and planning, and a deeper understanding of the processes underlying divergent economic, social and health trends. Spatial relationships include the physical distance between objects, the existence of a hierarchy of administrative areas within which data are collected, and characteristics such as contiguity or connectivity of places.
The team uses and develops analysis methods based on three different traditions:
multi-level regression procedures that take account of spatial hierarchies
deterministic procedures using Geographical Information Systems (GIS)
predictive models that apply spatial and temporal autocorrelation structures
In the past, statisticians have tended to deploy hierarchical models for estimation and autocorrelation models for the causal investigation of spatial processes. There is no reason why the approaches should not be integrated and we are now exploring this, mainly in the context of small area estimation. In the future we plan to develop guidance on the closely related issues of data smoothing (separating underlying patterns from random noise) and spatial methods for explanatory and causal analysis. Our goal is to understand the underlying stochastic processes inherent in these methods and to make use of the most powerful statistical methods where possible.
The aim of the programme is to carry out methodological research to develop methods for spatial analysis and modelling, focusing presently on implementing these methods to produce small area estimates. We also provide expertise, advice and quality assurance in aspects of spatial analysis and modelling and produce prototype computer programs to enhance aspects of analysis.
Current and planned areas of work
The Implementation Unit produces specific small area estimates, maintains a database of relevant survey, census and administrative data, and is also investigating model diagnostics. The unit has produced estimates of International Labour Organisation (ILO) unemployment (at local authority district level) and household income (at ward level). The team is currently updating these, and also investigating the possibilities of producing local estimates on other topics, such as crime.
The Research Unit developed the small area estimation methodology that we now apply and is exploring other spatial methods. Current projects include:
the multinational EURAREA project which is studying small area estimation techniques, their enhancement and empirical evaluation
Neighbourhood Statistics Service research: includes methods for recasting population characteristics onto alternative geographies and methodological aspects of designing and maintaining super output areas
methods for extending hierarchical estimates to alternative geographic bases
improving the procedures we use to fit spatial models