ABSTRACT
The most common reason for performing a regression analysis in epidemiology is to obtain estimates of the coefficients associated with the variables of interest, for example, the effect of an increase in particulate matter air pollution on the risk of respiratory death. In order to perform such analyses there will be a need for accurate estimates of exposures on which to base the associations with health. Often there will be locations and periods of time for which such data will not be available. This may be due to a fault in monitoring equipment or may be due to design. In many epidemiological studies, the locations and times of exposure measurements and health assessments do not match, in part because the health and exposure data will have arisen from completely different data sources and not as the result of a carefully designed study. This is termed the ‘change of support problem’ (Gelfand, Zhu, & Carlin, 2001). In such cases, a direct comparison of the exposure and health outcome is often not possible without an underlying model to align the two in the spatial and temporal domains. In this chapter we consider how predictions from exposure models, which are covered in detail in Chapters 9, 10 and 11 can be used in models for estimating health risks.
