07-11-2013 01:41 PM
I'm trying to use proc variogram to study spatial correlation in a variable (present2012). The difficulty is that the range class 0 and 2 through 50 make sense. The range class 1, however, is not representative, for some reasons I won't go into here. I know how to condition the variogram estimate on the lags from 2 to 50:
proc variogram data=summarizedbycentroid plots=all;
store out=storedmodel / label='WNV cases';
compute lagd=.001 maxlag=50;
coord xc=x yc=y;
model form=exp rangelag=2 to 50;
but I can't figure out how to do something similar for (0, 2:50). I've tried everything, but I'm pretty sure it's not supported. Is there some way I can run the variogram, save the pairs, delete those in class 1, then run the variogram with this modified set? Etc.
Thanks for any insights,
07-11-2013 02:02 PM
How about :
compute lagdistance=2 maxlags=25;
The implied LAGTOLERANCE would be 1. You could make it less if you wanted to exclude lag=1 pairs.