by Didier Renard » Sun Mar 15, 2020 4:32 pm
Hello
Up to my knowledge, no change has been brought to this procedure. But difficult to say if something has been modified or not in the internal functions. Moreover, you must remember that the model.auto() function corresponds to an iterative procedure; hence the result nay vary if the number of iterations is modified or if a tolerance is modified.
Could you be more specific and possibly send a view of the old and the new fits. As a matter of fact, the same experimental variogram can be fitted with two different models with two scores which may make these two trials undistinguished. If you are able to run the two tests again, you should run with the verbose option in order to trace the trajectory of both fitting procedures.
Finally it is essential to know if you consider a simple monovariate case with a single direction (the simplest case) or a multivariable multidirectional one.
Thanks for your help.