THE USE OF OPTIMAL ESTIMATION FOR GROSS ERROR DETECTION IN DATABASES OF SPATIALLY CORRELATED DATA

Abstract

For data, which are associated with a spatial position (location), it is very often so that data are spatially correlated. The distance plays the role time does in time series, while the directional dependence often is small or may be disregarded. This may be used to detect fross errors, using tools developed for optimal estimation in stochastic processes. Here methods like optimal linear prediction makes it possible also to estimate the error of prediction. A comparison of the difference between the observed and the predicted value with the error estimate, may then be used to identify a possible gross error