Hybridizing Alternative Models of an Epidemic Diffusion Process

Abstract

Seven alternative models of an epidemic diffusion model are compared using a binary autocorrelation statistic. The study area consisted of 28 local areas within an English county reporting measles notifications over a 222-week period. Each week's pattern of outbreak was converted to a binary code and plotted on seven graphs, each replicating a presumed spatial pattern of spread. Application of join-count statistics to the graphs allowed identification of the more plausible hypotheses. Results suggested that different spatial processes were important at different stages of an epidemic's development. A combined process, hybridizing several individual spatial processes, is put forward.