In the Getting Started section of the chapter on the VARIOGRAM procedure, a particular variogram is chosen for the coal seam thickness data. The chosen variogram is Gaussian with a scale (sill) of c = 7 . 5, and a range of a = 30. This choice of the variogram is based on a visual fit ”a comparison of the plots of the regular and robust sample variograms and the Gaussian variogram for various scale (sill) and range values.
Another possible choice of model is the spherical variogram with the same scale (sill) of c = 7 . 5 but with a range of a = 60. This choice of range is again based on a visual fit; while not as good as the Gaussian model, the fit is reasonable.
It is generally held that spatial prediction is robust against model specification, while the standard error computation is not so robust.
This example investigates the effect of using these different models on the prediction and associated standard errors.
data thick; input east north thick @@; datalines; 0.7 59.6 34.1 2.1 82.7 42.2 4.7 75.1 39.5 4.8 52.8 34.3 5.9 67.1 37.0 6.0 35.7 35.9 6.4 33.7 36.4 7.0 46.7 34.6 8.2 40.1 35.4 13.3 0.6 44.7 13.3 68.2 37.8 13.4 31.3 37.8 17.8 6.9 43.9 20.1 66.3 37.7 22.7 87.6 42.8 23.0 93.9 43.6 24.3 73.0 39.3 24.8 15.1 42.3 24.8 26.3 39.7 26.4 58.0 36.9 26.9 65.0 37.8 27.7 83.3 41.8 27.9 90.8 43.3 29.1 47.9 36.7 29.5 89.4 43.0 30.1 6.1 43.6 30.8 12.1 42.8 32.7 40.2 37.5 34.8 8.1 43.3 35.3 32.0 38.8 37.0 70.3 39.2 38.2 77.9 40.7 38.9 23.3 40.5 39.4 82.5 41.4 43.0 4.7 43.3 43.7 7.6 43.1 46.4 84.1 41.5 46.7 10.6 42.6 49.9 22.1 40.7 51.0 88.8 42.0 52.8 68.9 39.3 52.9 32.7 39.2 55.5 92.9 42.2 56.0 1.6 42.7 60.6 75.2 40.1 62.1 26.6 40.1 63.0 12.7 41.8 69.0 75.6 40.1 70.5 83.7 40.9 70.9 11.0 41.7 71.5 29.5 39.8 78.1 45.5 38.7 78.2 9.1 41.7 78.4 20.0 40.8 80.5 55.9 38.7 81.1 51.0 38.6 83.8 7.9 41.6 84.5 11.0 41.5 85.2 67.3 39.4 85.5 73.0 39.8 86.7 70.4 39.6 87.2 55.7 38.8 88.1 0.0 41.6 88.4 12.1 41.3 88.4 99.6 41.2 88.8 82.9 40.5 88.9 6.2 41.5 90.6 7.0 41.5 90.7 49.6 38.9 91.5 55.4 39.0 92.9 46.8 39.1 93.4 70.9 39.7 94.8 71.5 39.7 96.2 84.3 40.3 98.2 58.2 39.5 ; /*- Run KRIGE2D on original Gaussian model ------------ */ proc krige2d data=thick outest=est1; pred var=thick r=60; model scale=7.5 range=30 form=gauss; coord xc=east yc=north; grid x=0 to 100 by 10 y=0 to 100 by 10; run; /*- Run KRIGE2D using Spherical Model, modified range -*/ proc krige2d data=thick outest=est2; pred var=thick r=60; model scale=7.5 range=60 form=spherical; coord xc=east yc=north; grid x=0 to 100 by 10 y=0 to 100 by 10; run; data compare ; merge est1(rename=(estimate=g_est stderr=g_std)) est2(rename=(estimate=s_est stderr=s_std)); est_dif=g_est-s_est; std_dif=g_std-s_std; run; proc print data=compare; title 'Comparison of Gaussian and Spherical Models'; title2 'Differences of Estimates and Standard Errors'; var gxc gyc npoints g_est s_est est_dif g_std s_std std_dif; run;
Comparison of Gaussian and Spherical Models Differences of Estimates and Standard Errors: First 50 Observations Obs GXC GYC NPOINTS g_est s_est est_dif g_std s_std std_dif 1 0 0 23 43.9408 42.6700 1.27087 0.68260 2.05947 1.37687 2 0 10 28 41.6828 41.6780 0.00483 0.55909 2.03464 1.47554 3 0 20 31 38.9601 39.7285 0.76843 0.30185 1.93478 1.63293 4 0 30 32 36.1701 37.3275 1.15739 0.12705 1.54844 1.42139 5 0 40 39 33.8376 35.4320 1.59440 0.04872 1.37821 1.32949 6 0 50 38 32.8375 34.3930 1.55550 0.02983 1.22584 1.19602 7 0 60 35 33.9576 34.3155 0.35785 0.00195 0.54122 0.53927 8 0 70 30 36.9502 37.6669 0.71673 0.04006 1.20451 1.16444 9 0 80 31 41.1097 41.1016 0.00812 0.04705 0.99544 0.94839 10 0 90 28 43.6671 42.5216 1.14546 0.10236 1.57357 1.47121 11 0 100 23 41.9443 42.6511 0.70681 0.53646 2.20792 1.67146 12 10 0 25 44.6795 44.1959 0.48355 0.07833 1.09743 1.01910 13 10 10 31 42.8397 42.7496 0.09008 0.10982 1.46686 1.35703 14 10 20 34 40.3120 40.3634 0.05140 0.05315 1.54889 1.49574 15 10 30 39 37.7593 37.7648 0.00544 0.00889 0.94136 0.93247 16 10 40 44 35.6365 35.5471 0.08940 0.00595 0.75920 0.75325 17 10 50 44 35.0603 34.7042 0.35612 0.01564 1.05033 1.03469 18 10 60 41 36.0716 35.4737 0.59794 0.01321 1.18277 1.16957 19 10 70 36 38.1196 38.1040 0.01565 0.00315 0.89157 0.88842 20 10 80 33 41.2799 41.0734 0.20644 0.02446 1.22772 1.20326 21 10 90 30 43.2193 42.8904 0.32890 0.05988 1.49438 1.43450 22 10 100 26 41.0358 43.1350 2.09918 0.19050 1.93434 1.74384 23 20 0 29 44.4890 44.4359 0.05317 0.06179 1.23618 1.17439 24 20 10 35 43.3391 43.2938 0.04531 0.00526 0.95512 0.94986 25 20 20 39 41.1293 40.9885 0.14079 0.00675 1.18544 1.17870 26 20 30 43 38.6060 38.5300 0.07598 0.00898 1.08973 1.08075 27 20 40 49 36.5013 36.5275 0.02623 0.03037 1.33620 1.30583 28 20 50 49 36.1158 35.7959 0.31990 0.02535 1.31986 1.29451 29 20 60 49 36.8115 36.5397 0.27182 0.00835 1.11490 1.10656 30 20 70 39 38.4308 38.5182 0.08746 0.00257 0.89419 0.89162 31 20 80 36 41.0601 41.0449 0.01511 0.00766 1.18548 1.17781 32 20 90 33 43.1788 43.1073 0.07144 0.00613 0.94924 0.94311 33 20 100 27 42.7757 43.4689 0.69313 0.06770 1.52094 1.45324 34 30 0 35 43.3601 43.9579 0.59779 0.04662 1.32306 1.27644 35 30 10 39 43.1539 43.1448 0.00912 0.00245 0.72413 0.72167 36 30 20 44 41.2400 41.2166 0.02336 0.00528 1.10234 1.09706 37 30 30 52 38.9296 39.0178 0.08816 0.00489 1.04501 1.04012 38 30 40 57 37.2813 37.3412 0.05992 0.00804 0.89242 0.88438 39 30 50 57 36.7198 36.7558 0.03597 0.00652 0.83517 0.82865 40 30 60 55 37.2047 37.3407 0.13597 0.00682 1.00330 0.99648 41 30 70 48 38.8856 38.8919 0.00628 0.00285 1.01430 1.01145 42 30 80 43 41.0627 41.0663 0.00359 0.00260 0.97336 0.97077 43 30 90 36 43.0969 43.0465 0.05038 0.00194 0.51312 0.51118 44 30 100 29 44.5840 43.3474 1.23663 0.13593 1.57267 1.43674 45 40 0 36 42.8186 43.5157 0.69706 0.01976 1.25689 1.23713 46 40 10 40 42.8970 42.9168 0.01984 0.00301 0.95163 0.94862 47 40 20 52 41.1025 41.1824 0.07989 0.00193 0.96204 0.96012 48 40 30 60 39.3288 39.2992 0.02960 0.00451 1.05561 1.05111 49 40 40 67 38.2096 37.9680 0.24161 0.01791 1.29139 1.27349 50 40 50 68 37.3139 37.5055 0.19150 0.04039 1.51095 1.47056
The predicted values at each of the grid locations do not differ greatly for the two variogram models. However, the standard error of prediction for the spherical model is substantially larger than the Gaussian model.