Space -time clustering revisited

The history of space-time clustering concepts and methods are re- viewed briefly. The space-time clustering model of Ederer et al is investigated in detail. This method has been used extensively in the epidemiological literature, but we show that the distribution of main test statistic involved does not follow the distribution proposed by the authors. We note, too, that the two indices proposed are not statistically independent, leading to potential over-reporting in the epidemiological literature. We obtain the correlation between the original clustering indices and suggest a new combined test statistic which has the correct null distribution. We develop a fuller spatial model and illustrate the methodology using data from a study of the incidence of childhood leukaemia in Northern Ireland.