DETECTION OF OVERALL SPACE-TIME CLUSTERING IN A NONUNIFORMLY DISTRIBUTED POPULATION

We developed a test statistic based on an approach of Whittemore et al . (1987) to detect space-time clustering for non-infectious diseases. We extended the spatial test of Whittemore et al. by deriving conditio nal probabilities for Poisson distributed random variables. To combine spatial and time distances we defined a distance matrix D, where d(ij ) is the distance between the ith and jth cell in a three-dimensional space-time grid. Spatial and temporal components are controlled by a w eight. By altering the weight, both marginal tests and the intermediat e test can be reached. Allowing a continuum from a pure spatial to a p ure temporal test, the best result will be gained by trying different weights, because the occurrence of a disease might show some temporal and some spatial tendency to cluster. We examined the behaviour of the test statistic by simulating different distributions for cases and th e population. The test was applied to the incidence data of insulin-de pendent diabetes mellitus in Finland. This test could be used in the a nalysis of data which are localized according to map co-ordinates, by addresses or postcodes. This information is important when using the G eographical Information System (GIS) technology to compute the pairwis e distances needed for the proposed test.