A comparison of variogram estimation with covariogram estimation
Consider teh class of intinsically stationary spatial processes, which contains the class of second-order stationary processes. A measure of spatial dependence in the larger class is the variogram, from which optimal linear predictors can be constructed. For processes that are second-order stationary, these optimal linear predictors can also be expressed in terms of the covariogram. Traditionally, time-series forecasting has used the covariogram, but use of the variogram allows more general processes to be considered. These measures of spatial dependence are often unknown and have to be estimated from the data. In this article, we show that estimation of the variogram has important advantages over estimation of the covariogram.