1979 年 34 巻 3 号 p. 512-520
In geographic epidemiology, distributions of the categorized mortality or morbidity rates are visualized on maps. Visual study, however, by no means proves whether geographic aggregations occur by chance alone. We have developed a test of significance to assess the deviation from chance expectations of geographic patterns actually observed and have described it in this paper. The significance is tested by comparing the observed number of adjacent areas having concordant category pairs with the number obtained by a simulation based on the Monte Carlo method.
The test was applied to illustrations of geographic patterns of the categorized mortality from esophageal, stomach, and lung cancers in Japan, 1969-1971. For esophageal cancer, high mortality was significantly clustered in both sexes, and low mortality in males. For stomach cancer, both high and low mortality was significantly aggregated in both sexes. Low lung cancer mortality showed significant geographic aggregations in males.
Some prerequisites for the proper use of the test, the characteristics of the random numbers generated, the changeable significance levels, and the methodological alternative are discussed. The test is applicable not only to geographic epidemiology, but also to other scientific fields where an investigator is concerned with geographic clusters of variables or events.