Association testing
Examination of the intercorrelation of predictor variables indicated
that many of the morphological variables related to body size attributes
were highly correlated, however, none had a significant correlation
above 0.95 (Table S3 – S8), and therefore all were retained for
association analysis. Significant intercorrelations among traits were
consistent across samples and using these correlations we categorized
traits into the following four main groups: 1) body size, 2) female
sexual maturity, 3) larval growth, and 4) swimming ability. The “Body
size” category included the body metrics of length, weight, girth, and
interdorsal, which were all significantly positively correlated across
datasets (Table S3 – S8). This trait category also included migration
distance which was positively correlated with increasing body size
metrics, and migration timing which was negatively correlated with these
body size metrics. The male gonad mass was intended to serve as a proxy
for male sexual maturity, but it also was significantly positively
correlated with other body size metrics. However, the female gonad
metric was not correlated with the other body size metrics and was
potentially an accurate proxy for “female sexual maturity”. The
“larval growth” category was populated by the only two measures of
growth in the common garden experiment. The “swimming ability”
category contained the three metrics of swimming performance which were
all significantly positively correlated, and included migration day,
which was positively correlated.
We examined the relative strength of associations of the total 302 SNPs
with the traits within each of the four trait categories (Table S9), but
were primarily interested in associations of SNPs on the four
chromosomes chr01, chr02, chr04 and chr22, where evidence for range-wide
adaptive divergence was concentrated.