Spatial Correlation in Ecological Analysis

Abstract

This paper presents a statistical approach, originally developed for mapping disease risk, to ecological regression analysis in the presence of spatial autocorrelated extra-Poisson variation. An insight into the effect of allowing for spatial autocorrelation on the relationship between disease rates and explanatory variables is given. Examples based on cancer frequency in Scotland and Sardinia are used to illustrate the interpretation of regression coefficient and further methodological issues.

Higher incidence of childhood-onset type 1 diabetes mellitus in remote areas: a UK regional small-area analysis

“…Additionally, the measure of remoteness adopted was objective, based upon road distances, and thus has advantages over previously used measures such as urban-rural classifications, which are based upon more subjective judgements and do not necessarily reflect remoteness. Also our findings were little altered after adjustment for spatially correlated extra-Poisson variation, using Bayesian analysis [12], which if present could incorrectly exaggerate the significance of associations. Few previous ecological analyses of type 1 diabetes mellitus incidence rates have made such adjustments [5,7].…”

“…The regression coefficients for area characteristics in this model are adjusted to allow for spatially correlated rates, potentially caused by unmeasured explanatory variables. The use of this model for ecological analysis has been described in detail previously [12].…”

Population mixing and leukaemia in young people around the La Hague nuclear waste reprocessing plant

“…Poisson regression with random effects allows a closer approximation to the geographical structure of the data. Random effects may be considered as surrogates for unknown or unmeasured area-level covariates Clayton et al, 1993;Mollié, 1996). If the unobserved risk factors exhibit spatial correlation, it may result in the non-independence of the true incidence rate ratio between contiguous areas.…”

“…where e i is a random effect allowing for extra-Poisson variation Clayton et al, 1993;Mollié, 1996). The random effect e i was split into two components, the spatially unstructured extra-Poisson variation (heterogeneity), u i , and the spatially structured extra-Poisson variation (clustering), v i .…”

Spatial statistical methods in health

“…The latter may be present (e.g. see Clayton et al, 1993) arising, for example, through lesser variability of rates in neighboring densely populated urban areas as opposed to sparsely populated rural areas, or through an infectious aetiology of the disease. Such explicit spatial dependence may be incorporated into the model by including an additional spatially structured random effect term (e.g.…”