Spatial data analysis uncertainties introduced by selected sources of error

Abstract

Spatial data analysis uncertainty has been examined with various sources of error through simulation experiments. The general sources of the uncertainty are sampling error, measurement error, specification error, and location error. Location error is a unique error in spatial data analysis and occurs when an observed location deviates from its true location. We simulate spatial data analyses with different levels of location and measurement error, and compare the simulation results. Geographically aggregated pediatric blood lead level point data for Syracuse, New York, are utilized for the simulation together with a simultaneous autoregressive model. The results show that even with different levels of error, regression coefficients are quite robust. However, coefficient standard errors become larger with higher levels of location error and smaller administration units, such as census blocks.