Inheritance of Rest Period of Seeds in the Peanut
The classification of rest period given above produced highly
*skewed distributions not well suited to analysis of variance
tests. That method of classifying was called "equal interval"
because all class intervals were equal except the first, which
was of indeterminate breadth in the genetic sense. Transforma-
tions of the distributions were made by two other methods of
classification. It was necessary in making these transformations
to assume that the relative rank of F2 mean rest periods was
measured with reasonable accuracy. However, no doubt need
be entertained of significant regressions on that account since
failure to classify correctly could hardly produce spurious re-
gression.
The first transformation was to a rectangular distribution by
taking classes of equal frequency throughout. This provided
tests of regression on rank. The second transformation was an
attempt to normalize the distributions by taking class frequen-
cies proportional to areas of equal breadth under the normal
curve. It was done by dividing a range of 2.6 standard devia-
tions plus or minus into equal sections and calculating the pro-
portionate frequencies from Fisher's Table I (4). The total
range of 5.2 standard deviations centered at the mean includes
a little more than 99 percent of the total area under the normal
curve. Justification for this transformation lies in the supposi-
tion that distributions of rest period genotypes are probably not
far from normal type.
Most of the tests on rest period were made with the trans-
formed rectangular distributions but a few tests with the other
distributions are also shown in Table 12. When rest period
served as the dependent variable only the actual data were
used with no transformation, and no transformations of distribu-
tions of other variables were made in any case.
Values of F for the several tests of regressions on rest period
in cross 1 x 14 are presented in Table 12. The F values for
tests of regression on seed shape, seed coat color, yellow seed-
lings, and Valencia plant type of the 22 recorded characters
in the four crosses have been omitted. As shown by the column
headings each series of F values is listed in two columns, "Be-
tween" and "Within", to indicate which was the larger mean
square in the ratio F. A significantly greater between mean
square indicates that group means of the dependent variable are
less alike than expected with random sampling. The increment
of variance is attributed to association with the independent