Applied Bayesian Modelling, 2nd Edition

Chapter 8Models for spatial outcomes and geographical association

8.1 Introduction

Advances in spatial data analysis refer to a central core of knowledge but show many distinct features in the specialisms involved. Thus many Bayesian applications have occurred in spatial epidemiology, with methodologically oriented overviews including Pfeiffer et al. (2008), Waller and Gotway (2004), Beale et al. (2008), Graham et al. (2004), Schrödle and Held (2011), Auchincloss et al. (2012) and Jerrett et al. (2010). Here a major element is the assessment of spatial clustering of relative disease risk, often for irregular lattice systems (e.g. administrative areas). A more long-standing tradition of spatial modelling has occurred in spatial econometrics with Anselin (2006, 2010), Pace and LeSage (2010), Getis et al. (2004), Arbia and Baltagi (2009) and LeSage (2008) providing recent overviews, and with LeSage and Pace (2009) reviewing Bayesian principles in this area. Here the major emphasis lies in describing behavioural relationships by regression models, whether the data are defined over areas, or for individual actors (house purchasers, firms, etc.) involved in spatially defined behaviours. A third major specialism occurs in geostatistics, where a continuous spatial framework is adopted, and the goal is often to smooth or interpolate between observed readings (e.g. of mineral concentrations) at sampled locations (Diggle and Ribeiro, 2007; Gaetan and Guyon, 2009). Providing a common thread ...