Description
Data transformations (Log and Box-Cox) do not work when data values are less than or equal to 0. For empirical Bayesian kriging, the LOGEMPIRICAL transformation requires all data to be greater than 0.
Solution
A Normal score transformation can be used to obtain a dataset that has a Normal (Gaussian) distribution. For empirical Bayesian kriging, the EMPIRICAL transformation can be used with data less than or equal to 0.