The use of sampling weights in Bayesian hierarchical models for small area estimation
Empirical Bayes and Bayes hierarchical models have been used extensively for small area estimation. However, the sampling weights that are required to reflect complex surveys are rarely considered in these models. In this talk, Jon Wakefield will describe a method for incorporating the sampling weights for binary data when estimating, for example, small area proportions or predicting small area counts. He will consider spatial random effects models with computation carried out using a relatively new and fast technique. Simulation results will show that estimation of mean squared error can be reduced when compared with more standard approaches. Bias reduction occurs through the incorporation of sampling weights, with variance reduction being achieved through hierarchical smoothing. He will also analyze data taken from the Washington 2006 Behavioral Risk Factor Surveillance System. This is joint work with Cici Chen Bauer and Thomas Lumley.
Dr. Wakefield has worked extensively in the general area of spatial epidemiology and was formerly a member of the Small Area Health Statistics Unit (SAHSU) at Imperial College in London. He has particular interests in the sources and alleviation of ecological (or aggregation) bias and has worked on study designs for supplementing ecological information. His other interests include spatial-temporal models for infectious disease data, cluster detection, disease mapping, spatial regression, Bayesian methods in biostatistics and epidemiology, and genetic epidemiology. He has received the Guy medal in Bronze from the Royal Statistical Society, is a Fellow of the American Statistical Association, and is a former chair of the Department of Statistics at the University of Washington.