Simulation-based Assessment of a Geostatistical Approach for Estimation and Mapping of the Risk of Cancer

Abstract

This paper presents a geostatistical methodology that accounts for spatially varying population sizes and spatial patterns in the processing of cancer mortality data. The binomial cokriging approach is adapted to the situation where the variance of observed rates is smaller than expected under the binomial model, thereby avoiding negative estimates for the semivariogram of the risk. Simulation studies are conducted using lung cancer mortality rates measured over two different geographies: New England counties and US State Economic Areas. For both datasets and different spatial patterns for the risk (i.e. random, spatially structured with and without nugget effect) the proposed approach generally leads to more accurate risk estimates than traditional binomial cokriging, empirical Bayes smothers or local means.