Florida Agricultural Experiment Station
variable. A greater within mean square indicates that the group
means of the dependent variable are more alike than expected
with random sampling. Values of F for the points of 5 and 1
percent probability, Snedecor (15), are given at the bottom of
each column and all values of F in the table which exceed the
5 percent point have been set in bold face type.
The F values of tests of the regressions of the two generation
measures of rest period on each other are highly significant with
each of the three methods of classification when judged by the
criteria applicable with normal distributions. That they are
really significant seems probable from a comparison with other
F values in the table which range much smaller generally. The
F values of the other tests which were not listed were also
much smaller. Significant regressions of the different genera-
tions of rest period on each other were to be expected, since
strong heritability of the character was already demonstrated.
Perhaps the principal value of these tests was to provide F
values derived from highly skewed distributions with strong
regressions to compare with other F values also derived from
highly skewed distributions where the order of regression was
unknown.
Some of the characters other than rest period also had highly
skewed distributions. Seed coat color was classified sometimes
in ratios approximating 1:2:1 and in other cases the ratios
approximated 3:1. Valencia plant type has a highly skewed
distribution in cross 1 x 14, as may be seen in the first experi-
mental ratio of Table 11. This character is controlled by dupli-
cate pairs of genes with one pair coming from each parent of
the cross; thus it is hardly expected that genetic regressions
involving Valencia plant type should be found in F2. The order
of F values obtained with characters having skewed distribu-
tions appears not different from that of characters with distribu-
tions generally symmetrical.
Only a few of the F values obtained in tests involving two
separate characters exceed the points of significance and those
few to no great extent. Results of the several tests on any pair
of characters are frequently inconsistent or contradictory. It
might appear, e. g. in Table 12, that greater between mean
squares were obtained too frequently with both plant weight
and weight of nuts to be ascribed to chance but in the other
three crosses no significant F values were obtained and the