1-III_am

[34] Potential influence that use of sample data has on spatial analysis

- A case study with spatial autocorrelation analysis -

○Ikuho Yamada¹ (1. Center for Spatial Information Science, the University of Tokyo )

Keywords:sample data, spatial analysis, spatial autocorrelation, local Moran statistic

This study investigated potential influence that use of spatial data that were obtained through a sample survey would have on spatial analysis based on statistical simulations. In the simulations, spatial patterns with clusters of high attribute values were stochastically generated and analyzed using the population data as well as sample data so as to assess if and how results of the analyses with the two types of data would differ from each other. The spatial analysis method used here was the local Moran statistic. Results suggested that occurrence of type I errors was not much affected by the difference of the data types. On the other hand, the statistical power of cluster detection was found to decrease as the sampling rate of the sample data decreased. In addition, this tendency was more critical near the border of the simulated clusters, implying that shapes of the clusters were difficult to accurately determine with the sample data.