Patterns of variance and covariance for specific traits
Suspensorium length, palatine height, and ascending process length have
strong positive correlation with PC axis 1 (0.75,0.75, 0.71
respectively) while pelvic girdle length, dentigerous arm depth and
lower jaw length have strong negative correlations (-0.79, -0.74, -0.72
respectively; Appendix Table E5). PC2 is defined by positive
correlations with ectopterygoid (0.45) and jaw opening in-lever (0.22)
and strong negative correlations with head depth (-0.85), dentigerous
arm width (-0.77), and cranial height (-0.76; Appendix Table E5). Most
variation across both axes can be attributed to patterns in the Little
Lake P matrix (93% of variance across PC1, and 63% variance across
PC2), however, we can still identify clear differences in the
relationship between traits for each population. For instance, there is
a strong positive relationship between covariation patterns of lower jaw
length and jaw opening in-lever in Little Lake, but this relationship is
non-existent in Crescent Pond (Figure 8; Appendix Table E6). In fact,
many relationships between traits captured by PC1 are strong in Little
Lake but weak in Crescent Pond. Maxillary head height is the only trait
in Crescent Pond to show a strong positive relationship with head depth,
dentigerous arm width, and cranial height, while several traits in
Little Lake share this relationship (Figure 8).
Finally, we found that overall variation within traits was similar
between Crescent Pond and Little Lake (T test: t = -1.97, df = 27.45,
p-value = 0.059) but variances in Little Lake were greater than
variances in Crescent Pond for all traits. Although similar between
populations, only the variation in Crescent Pond met null expectations
associated with the laws of independent assortment and segregation
(expected variation: 0.14, observed variation: 0.013; T test: t = -0.14,
df = 33.72, p-value = 0.89). Variation in Little Lake was significantly
higher than expected (expected variation: 0.007, observed variation:
0.029; T test: t = 3.05, df = 21.37, p-value = 0.006). We found that
covariation estimates between traits were not significantly different
from one another (LM: F=1.39, df=17, p-value=0.13), but that overall
Little Lake had higher estimates of covariation than Crescent Pond (LM:
F=30.1, df=17, p-value<0.01).
The interaction between population and trait significantly affected
regression coefficient values (LM: F= 1.93, df=17, p-value=0.013).
Within Crescent Pond, head depth, maxilla length, and palatine height
had significantly higher regression coefficients than 50% of measured
traits. Within Little Lake, head depth, lower jaw length, and
dentigerous arm width had higher regression coefficients than 50% of
measured traits. Comparing traits between populations regression
coefficients for orbit diameter (Tukey’s HSD: t ratio=2.67,
p-value=0.0078), palatine height (Tukey’s HSD: t ratio=3.45,
p-value=0.0078), jaw closing in-lever (Tukey’s HSD: t ratio=2.22,
p-value=0.027), and maxilla length (Tukey’s HSD: t ratio=2.18,
p-value=0.003) were all higher in Crescent Pond than in Little Lake.
Lastly, we found that squared correlation coefficients significantly
varied between traits (LM: F=10.19, df=17, p-value<0.01) where
dentigerous arm width, maxilla length, lower jaw length, and palatine
height, had higher correlation coefficients than 50% of traits in the
data set.