[Additional limitations of the Shrikhande method and their verification].

Shrikhande's method was checked on linear-variety common wheat Lutescens 62. Two causes for a distortion of the ratio of genotypic and paratypic components of the variance were established: parameter b of Smith's law is not constant for a given field and depends on the method of parcelling: Smith's law is not confirmed at all. A two-stage test of Shrikhande's method correctness is suggested for checking its limitations in practice. The analysis of variance for any given parcelling is done at the first stage. Significant ratios of between-cluster mean squares and within-cluster mean squares indicate that Smith's law is confirmed. The second stage consists in the analysis of variance of lg V(X) linear regression by lg X or in the graphical analysis of the function lg V(X) = = f (lg X).