AQMeN - Applied Quantitative Methods Network

Alistair’s undergraduate and Masters degrees are both from The University of Edinburgh, in Geography and Geographical Information Systems (GIS) respectively. He did his PhD in Geography at the Pennsylvania State University. In between the latter degrees he worked at Lancaster University and then at the Macaulay Institute in Aberdeen.

Alistair’s research continues to involve GIS. As GIS is unlikely to be familiar to all in AQMeN it is worth saying a bit more about. Data on which geographic information is built consist in their most primitive form of two parts: a location, uniquely identified by a set of (x, y) coordinates, and one or more attributes, the latter recording some characteristics (i.e. variables) of interest for that location. These data primitives are the basis for modelling real-world entities in a GIS database. Crucially, other GIS functions enable the (x, y) coordinate information stored in the database to be used in differing ways. A key example of this would be where values from different data sets may be linked together where those data sets have locations in common.

Another key area associated with GIS is spatial data analysis, in which the core concern is to investigate how recorded values vary with location, and why. Frequently the geographical distribution of values is not random and instead exhibits spatial autocorrelation, whereby values for adjacent or nearby locations are more likely to be similar than values for locations further apart from one another. Spatial autocorrelation is problematic for conventional (non-spatial) statistical analysis because it violates the assumption of independence. In contrast, much spatial data analysis focuses on exploring, describing and modelling spatial dependence.

As well as spatial autocorrelation, spatial data analysis spans other well-known challenges. Challenges posed by ‘census geography’ have been a main focus in Alistair's own work. Primary issues with census geography concern the fact that boundaries of census reporting areas are typically defined arbitrarily, at differing levels of aggregation, and have been modified from one census. In turn this affects the patterns, relationships and changes observed using census results. Alistair’s research on census geography has focussed in particular on understanding the effects census area definition has on regression models, on the potential to develop and apply optimised regionalisations of census data, and on spatial dis-aggregation methods.

At Dundee, Alistair teaches modules in quantitative methods and in GIS. As well as undergraduate level he also teaches on the Masters programmes in Applied Population Geography and Social Research Methods. A selection of materials Alistair uses in this teaching may be of interest to those seeking further information.

Texts
• Fotheringham, Brundson and Charlton (2002) Geographically Weighted Regression: The Analysis of Spatially Varying Relationships, Wiley
• Goodchild and Janelle (2004) Spatially Integrated Social Science, Oxford
• O’Sullivan and Unwin (2003) Geographic Information Analysis, Wiley
• Rees, Martin, Williamson (eds) 2002, The Census Data System, Wiley
• Wilson, J. and Fotheringham (eds) (2008) The Handbook of Geographic Information Science, Blackwell

Web-based resources of UK relevance
• GEO-REFER project (Geographical Referencing Resources for Social Scientists) – http://www.restore.ac.uk/resources/grefer.php
• University College London Centre for Advanced Spatial Analysis – http://www.casa.ucl.ac.uk/
• The Office for National Statistics ‘Beginners Guide to UK Geography’ – http://www.statistics.gov.uk/geography/beginners_guide.asp

USA-based initiatives
• Centre for Spatially Integrated Social Science – http://www.csiss.org/
• University Consortium for Geographic Information Science – http://www.ucgis.org/

International
• Center for International Earth Science Information Network, Socioeconomic Data and Applications Center – http://sedac.ciesin.columbia.edu/

GIS / spatial data analysis software:
• GeoDa (and more recent software) – http://geodacenter.asu.edu/software
• Software for Geographically Weighted Regression – available for purchase from http://ncg.nuim.ie/ncg/GWR/software.htm