Evolutionary Developmental Biology (1): The Evolution of Ontogeny - Geometric Morphometrics for Biologists - page 323

FIGURE 11.13 Principal components analysis of the traditional measurements, pooling the ontogenetic series
of S. gouldingi and S. manueli.

FIGURE 11.14 Principal components analysis of the traditional measurements, pooling the ontogenetic series
of cotton rat and house mouse.

allometric coefficients of traditional data average 0.981, corresponding to an angle of
11.13 . Thus, only correlations higher than 0.987 or angles smaller than 9.10 indicate any
greater similarity than expected by chance. The correlation obtained from the data is
indeed higher than 0.987, but that value of 0.994 no longer seems remarkably high. For the
two rodent species, the average correlation obtained by randomly permuting the allome-
tric coefficients of traditional data is 0.89, corresponding to an angle of 27 . So, in this case,
only correlations higher than 0.96, or angles smaller than 16.3 , indicate any greater
similarity than expected by chance. The observed correlation of 0.984 is higher than 0.96,
but again, 0.984 no longer appears to be impressive evidence of a conserved ontogeny.
In striking contrast, in analyses based on geometric data, the average correlation between
randomly permuted vectors is

0.002, corresponding to an angle of 89.9 ; the value that
we would expect for randomly related vectors.

2