Theoretical Semivariogram Model Fitting - Use of "Lag 0"

CThickstun — Wed, 22 May 2024 20:41:16 GMT

I was hoping I could get some clarification on how SAS 9.4 handles the semivariance estimate for Lag 0 in the semivariogram model fitting process. A visual inspection of various fitted variogram curves for my data suggest that SAS is either ignoring or weighting down the importance of the Lag 0 semivariance estimates in fitting the semivariogram curves. Given that Lag 0 generally contains substantially fewer data points than the other lags under consideration, this seems reasonable, but I have been unable to find a description in the SAS User's Guide that details exactly how this process is handled.

Am I interpreting The VARIOGRAM Procedure - Quality of Fit page correctly to mean that Lag 0 is ignored in the fitting process when it says: "the weights wi2 are taken at i = 1, . . . , k." or am I missing something in the documentation elsewhere?

I am using this general code form for my model fitting:

PROC VARIOGRAM DATA = LIB.DATA  PLOTS(ONLY) = FIT;
COMPUTE AUTOCORRELATION
	LAGD = x
	MAXLAG = y
	CL ROBUST 
        AUTOCORR(WEIGHTS = DISTANCE (NORMAL)
		;
COORDINATES XC = X_CORD YC = Y_CORD;
MODEL FORM = AUTO (MLIST = (EXP, SPH, GAU, SHE) NEST = 1) 
	CL FIT = ROBUST
	CHOOSE = (STATUS SSE);
	VAR RESID;
RUN;

topic Theoretical Semivariogram Model Fitting - Use of "Lag 0" in Statistical Procedures

Theoretical Semivariogram Model Fitting - Use of "Lag 0"