07-10-2013 03:47 PM
I used proc reg to regress yield on spatial coordinates in order to detrend the data prior to empirical semivariogram in proc variogram. The SAS script is as follows
easq=easting*easting; /* quadratic trend term */
eastnort=easting*northing; /* interaction of easting and northing */
/* Trend surface analysis */
model yield=easting northing easq norsq eastnort/scorr1(tests) partialr2;
output out=detrend r=r p=p;
Each of the terms in the model seems to be significant different from zero which implies significant influence on yield. Does it mean there is trend in the data? IF yes, I can proceed to fit the empirical semivariogram on the residual of the output file "detrend". Please let me know if I am right, please.
Initially, I could not figure out the presence of trend by visualizing the spatial distribution of yield in 2-dimensional space of northing and easting due to large size of dataset. Could anyone suggest ways of visualizing such a big dataset of over 12,000 in geographical space? Please see output from above analysis for comment on whether detrending the data is good.