The analysis of putative environmental pollution gradients in spatially correlated epidemiological data

The analysis of putative sources of health hazard has recently received considerable attention. Most analyses, however, assume that the observations - such as counts in census tracts or exact locations of cases - are not spatially correlated. This assumption, if violated, could have serious consequences for inferences made from such data. This paper outlines a number of simple Monte Carlo test procedures for this situation.