Accuracy of Count Data Estimated by the Point - in - Polygon Method Yukio

This paper analyzed the accuracy of count data estimated by the point-in-polygon method. A point-in-polygon interpolation model was proposed which was based on a stochastic distribution of points and the target zone, in order to represent a variety of situations. The accuracy of estimates was numerically investigated in relation to the size of the target zone and the distribution of points, and the optimal location of representative points was discussed. The major findings obtained in this paper were as follows: 1) though the relative accuracy of estimates generally increases monotonously with the size of the target zone, the monotoneity is often disturbed by the periodicity in the spatial configuration of source zones and the point distribution; 2) the point-in-polygon and the areal weighting interpolation methods have the same accuracy of estimates when points are concentrated in less than 12-15% area around the representative point in source zones; 3) the point-in-polygon method is not so robust against the locational gap between points and the representative point; 4) the optimal location of representative points is given by the spatial median of points.