Fig. 3.

Bias of the standard error of difference (SED) and empirical Type I errors of the *t*-test and *F* test for a nominal error of 5% for the 1954 hemp, beet, and 1958 hemp crops using Designs 4, 9, 12, and 13 and all analyzed models: B, basic; *r*, random walk; P and *p*, power; S and *s*, spherical; G and *g*, Gaussian; A, anisotropic power model; b* indicates the best model.