[Evaluating the significance of differences between variants].

Experimental results of multiple comparisons for Tukey's, Newman-Kuels', Duncans' tests and for the t-test are compared. When the number of variants is more than two, the t-test is shown to reduce the value of the least significant difference, so that its use for multiple comparisons is incorrect. Duncan's multiple range test tends to increase the value of significant differences. The adequacy of Tukey's and Newman-Keuls' tests to the F-test is established.