Data Analysis
We generated a phylogenetic tree for our study species to help control
for any potential phylogenetic non-independence in our dataset before
testing our hypotheses H1-H4. We created a phylogeny from the trimmed
R20120829 mega-tree (Gastauer & Meira-Neto 2016) for our study species
using the ‘brranching’ R package’s ‘phylomatic’ function (Chamberlain
2018). We assigned fossil-calibrated branch lengths for our tree using
the ph_bladj function in the ‘phylocomr’ R package (Ooms & Chamberlain
2018). Any polytomies were randomly resolved using the ‘multi2di’
function in the ‘ape’ R package (Paradis & Schliep 2018).
We used our phylogenetic tree to compute a phylogenetic
variance-covariance (VCV) matrix to test each of our four a priori
hypotheses. Our VCV was calculated using the ‘phytools’ R package and
its ‘phyl.vcv’ function (Revell 2012). This approach assumes traits (in
our case the carbon assimilation and heat tolerance metrics) followed a
Brownian model of evolution and that trait variance was proportional to
branch lengths between two species and their most recent common
ancestor. We divided the product of the inverse VCV and the observed
trait values by the sum of the inverse VCV matrix to calculate the
ancestral trait value at the root of our phylogeny (Blomberg, Garland &
Ives 2003; Swenson 2014). These root trait values were used to calculate
a phylogenetically corrected covariation matrix among traits that was
rescaled to compute Pearson’s r using the ‘cov2cor’ function in R’s base
‘stats’ package (Core 2020). The t-statistic and an α = 0.05 were used
to test for significant trait correlations.
Below, we present the phylogenetically corrected correlations among
traits as phylogenetically independent contrasts (PIC) for graphical
purposes only. PICs were computed as the difference between two daughter
nodes standardized by the square root of the sum of branch lengths
(Felsenstein 1985), and performed using the ‘pic’ function in the ‘ape’
R package (Paradis & Schliep 2018) which results in n -1
contrasts where ‘n ’ is the number of species in the phylogeny.
The correlations and PICs are calculated differently, but provide
effectively analogous results. We present only the statistics from the
phylogenetically corrected correlations for simplicity in our figures.
All analyses were conducted using R version 4.0.2 (Core 2020).