The discipline of spatial statistics has traditionally been divided into three themes: geostatistics, discrete spatial models, and spatial point processes. Much of spatial epidemiology is concerned with discrete spatial models of the sort described in Chapters 6 and 7, with any residual spatial variation in risk or exposure (meaning variations beyond those caused by individual-level risk factors) being estimated at the level of health regions or census areas. Geostatistical models operate continuously in space, where the residual spatial component is diﬀerent at each and every location in the study area, but varying with some degree of smoothness. Two individuals located in close proximity to one another are assumed to have values of this residual risk that are quite similar, and similarity decreases as the distance of separation gets larger. Observational data on human health outcomes tend to be made available on spatially discrete scales, usually case counts in administrative regions, and geostatistics in epidemiology has generally been conﬁned to modelling environmental exposures. Study data, however, can often contain precise spatial information, such as full street addresses or GPS coordinates of the subjects’ homes, and geostatistical models are worth consideration when location data of this sort are available.