A statistical consideration on the mapping of mortality

The mortality map shows a spatial distribution of the death phenomena over a certain period of time and facilitates analysis of factors which influence disease. In constructing a mortality map, it is desirable from the geographical point of view to make each areal unit as small as possible so that we can compare it with other regions with different characters. On the other hand, since the number of deaths in each region decreases as each areal unit becomes smaller, the comparison of the level of mortality rate among the regions is known to become more difficult owing to the influence of random errors. Therefore the areal unit has to be fairly large from the statistical point of view. Those two demands conflict with each other. In practice it is necessary to set up appropriate areal units to make a good compromise between those two conflicting demands. In this paper, this problem is discussed in terms of statistical theory.