comparative method Flashcards
uses?
exploring trait evolution above species level
what is ancestral state of particular trait?
how fast a certain trait evolves within a phylogeny
why are birds so colourful
why has sociality evolved so many times
why differences in genome size
any time there is difference in a trait - can ask why they have evolved
comparative method balls example
hypothesis: sperm competition lead to larger testes in polygamous males
body mass vs testes mass regression of primate species
statistical assumption violated here:
-would assume that all data points are INDEPENDENT
-however here - closely related species have similar traits - including all these species risks violating this
-e.g. grouped all tamarins together because all are monogamous, having separate would violate
species as unit of replication?
comare two species
one longer lived and higher laydate heritability
another shorter lived and lower laydate heritability
cannot conclude that lifespan affects heritability
only comparing 2 species so any number of differences could be the cause
comparative tests no. of species = N
need adequate replication for tests to have statistical power (taxon = unit of replication)
2 statistical considerations?
- correlation =/= causation
need to be wary of effects of:
i. reverse-causation
ii. third variable problem - phylogenetic non-independence
form of pseudoreplication
have graph that looks like good correlation
however bottom half of graph is all one part of the phylogeny and the top half another
now no correlation
why are closely related species not individual data points?
phylogentic time lags
phylogenetic niche conservatism
shared adaptive responses
phylogenetic time lags
traits are not evolved/lost instantly, may persist for some time even when they no longer have a function
phylogenetic niche conservatism
daughter species tend to occupy same niche as their parent species
therefore may experience similar selective forces
shared adaptive responses
ease/difficulty with which a species can evolve or lose a new trait depends on the traits it already has
therefore relatives may respond similarly to selection as they have similar traits
dealing with phylogenetic non independence
take into account fact that phylogenetically related species are NOT independent data points
phylogenetic trees allow control for phylogeny and make statistically valid comparisons
INDEPENDENT CONTRASTS
independent contrasts
add to notes
independent trait evolution occurs ALONG BRANCHES
look at independent contrast of trait evolution along branch to A and the branch to B
POINT d1
two traits X and Y
take difference of each trait between A and B
do same for other terminal nodes C and D for trait X and Y
then infer state of ancestral node - average the two descendents (A and B) then use that value to compare with ancestral node of C and D
HOWEVER - longer branch lengths = expect to see more evolution along them
can standardise the contrast by taking the independent contrasts and dividing them by the sqrt of branch length
null model for how traits change?
brownian motion
shows longer period of time = more you’d expect the branch to diverge
why we need to standardise
variance covariance matrix
root to tip of tree = some value (time units)
example = 28 (bird order phylogeny)
put on variance covariance matrix how much time is shared between pairs of orders /28
used to correct in phylogenetic comparative methods
phylogenetic signal?
add to notes
not all traits as phylogenetically conserved as expected under brownian motion model (null model)
Pagel’s lambda - take variance-covariance matrix
multiply all off-diagonal elements by lamda (value 0-1)
can use maximum likelihood to identify lambda value that makes data most likely
-> can also imagine lambda as a branch length transformation
lambda = 1
strong phylo signal
consistent with what would be seen if traits evolves under brownian motion
lambda = 0
no phylo signal
phylo signal - phenomenon for related species to resemble each other, more than they resemble species drawn at random from a phylogenetic tree
why phylogenetic signal matters?
some traits have strong signal
(Body size)
others typically weak
(behavioural traits)
if trait has strong signal - might be able to predict the trait on phylogeny alone
comparative method for investigating why female birds are sometimes so brightly coloured
get bird types that:
male colourful, female drab
both colourful
both drab
2 hypotheses:
H1 - female coulouration only due to restraints - just correlated response with male colouration
H2 - female colouration is also under selection for their own signalling needs
quantified plumage colour
divided bird pictures into colour patches on commonly highly coloured areas
got RGB scores for each
plotted in 3D (RGB each axis)
for each patch in male and female birds - found 120 datapoints that were closest to that male or female patch colour
what proportion were male or female?
got MALENESS SCORE for colouration on males and females
got maleness score for each species
then used phylogenetuc mixed model comparative method
predicted variety of predictor values and accounted for them (whether they caused brighter colour or hindered it)
found high correlation between male and female colour (high maleness scores for female colours?)
females never really brighter than males
high correlation - when males get brighter - females do to
SUPPORT FOR H1 - constraint hypothesis
then saw how the life history predictors affected male and female colouration differently
-more colourful as body size bigger -
steeper in females
-stronger sexual selection = brighter males, drabber females - greater dichromatism - but not due to males brighter - due to females getting drabber
^ these factors showed that some instances they correlated (support for H1)
however other instances where responses were entirely independent - strong support for H2 - own signalling needs
“some female colouration driven by their own signalling needs”
