This really looks like "nothing happened".  The spatial correlation is still at 0.1, and that makes me think that the values of X and Y are such that the Euclidean distance between observations is huge, and consequently, no correlation is fit.  Ridging might help here, so try adding:   NLOPTIONS tech=nrridg;   as a line.   Also, the documentation says c-list has the names of the numeric variables (so check that X and Y aren't somehow still character values), for the ith vector and the jth vector, which correspond to the ith and jth observations.  That implies a dependence on the sort order of the observations, so it would probably be good to presort the data on X and Y.   And last, "The practical range of a (second-order stationary) spatial process is the distance at whcih the correlations fall below 0.05.  For the SP(EXP) structure this disance is 3*alpha." This alpha is the denominator in the covariance equation, so maybe by setting the initial guess to 0.1, it gets stuck.  You might try (hopefully I have these in the correct order):   parms (0.3) (0.1 to 0.9 by 0.1);   This should lead to a grid search across various correlation values, and maybe from there something will get "unstuck".   Steve Denham ... View more