make_unit_tree: Create unit.tree object

In mwpennell/arbutus: Evaluate the adequacy of continuous trait models

Create unit.tree object

Description

Generic method for creating a 'unit.tree' object

Usage

make_unit_tree(x, ...)

Arguments

x	fitted model object or phylo/multiPhylo object
...	additional arguments

Details

This function is a generic function which takes a fitted model object (from a number of different packages), rescales the phylogeny based on the model fitted parameters, computes the contrasts on the rescaled phylogeny and returns an object of class unit.tree. The unit.tree object can then be used to assess the adequacy of phylogenetic models of continuous character evolution. Alternatively, the function can take a phylo object or a multiPhylo object and compute the contrasts on the phylogeny without rescaling. This is only meaningful for downstream analyses if the phylogenies have been rescaled beforehand. Currently, the following object types have been implemented:

• a gfit object returned from fitting a model of continuous character evolution using fitContinuous in the geiger package.

• a fit.mle object returned from fitting a model of continuous character evolution using find.mle in the diversitree package.

• a mcmcsamples object returned from fitting a model of continuous character evolution using MCMC methods in the diversitree package. make_unit_tree will apply the same trait dataset to a set of unit trees based on sampled parameters. By default this will create a unit tree for every sample in the mcmc chain. To modify this, additional arguments can be use. burnin specifies how many samples to remove from the beginning of the chain. thin specifies the thinning interval (e.g. if thin=5, the function will create a unit tree from every fifth parameter set sampled. sample specifies how many samples to draw from the MCMC run.

• a gls object returned from fitting a phylogenetic generalized least squares model of character correlation using gls in the nlme package.

• a pgls object returned from fitting a phylogenetic generalized least squares model of character correlation using pgls in the caper package.

• a phylolm object returned from fitting a phylogenetic generalized linear model of character correlation using phylolm in the phylolm package. As the phylogeny is not returned with the phylolm object, a phy argument must also be specified.

• a phylo object. If a phylo object is supplied, the tree is assumed to have been rescaled previously. A second argument data must also be provided included the trait data as a named vector with names equal to the tip.labels of the phylogeny.

• a multiPhylo object. If a multiPhylo object is supplied, the tree is assumed to have been rescaled previously. A second argument data must also be provided included the trait data as a named vector with names equal to the tip.labels of the phylogenies. Note that this function will append the same data set to every tree in the multiPhylo object.

• a OUwie object returned from fitting a model of continuous character evolution using OUwie in the OUwie package.

Value

a unit.tree object containing (or a list of unit.tree objects, each containing) the following elements:

phy

a phylo object. If a model fitted object has been supplied (see details), the phylo object will be rescaled based on fitted model parameters.

data

the original comparative data

pics

a matrix consisting of the contrasts calculated using the original data on the (rescaled) phylogeny. The matrix has two columns: the "contrasts" and a "variance" for each contrast.

See Also

pic, arbutus

Examples

## finch data
data(finch)
phy <- finch$phy
dat <- finch$data[,"wingL"]

## using just the given phylogeny
## without rescaling
unit.tree.phy <- make_unit_tree(phy, data=dat)

## Not run: 
require(geiger)
## fit Brownian motion model
## using geiger's fitContinuous function
fit.bm <- fitContinuous(phy=phy, dat=data, model="BM",
                                 control=list(niter=10), ncores=1)

## this creates a 'gfit' object which can be used
## in 'make_unit_tree()'
unit.tree.geiger <- make_unit_tree(fit.bm)
unit.tree.geiger

require(diversitree)
## fit Brownian motion model using ML
## using diversitree's find.mle function

bmlik <- make.bm(phy, data)
fit.bm.dt <- find.mle(bmlik, 0.1)

## this creates a 'fit.mle' object which can be used
## in 'make_unit_tree()'
unit.tree.dt <- make_unit_tree(fit.bm.dt)

## fit Brownian motion model using MCMC
mcmc.bm.dt <- mcmc(bmlik, x.init=1, nsteps=1000, w=1)

## construct a unit tree object from mcmcsamples
## removing 100 samples as burnin
## and sampling 10 parameter sets
unit.tree.mcmc <- make_unit_tree(mcmc.bm.dt,
                       burnin=100, samples=10)


require(nlme)
## Use pgls to look for a correlation between two traits

t1 <- data
t2 <- td$data[,"tarsusL"]
dd <- cbind.data.frame(t1, t2)

## fit gls model with corPagel correlation structure
fit.gls <- gls(t1~t2, data=dd, correlation=corPagel(phy=phy, value=1))

## this creates a 'gls' object which can be used
## in 'make_unit_tree()'
unit.tree.gls <- make_unit_tree(fit.gls)


## End(Not run)

mwpennell/arbutus documentation built on Oct. 6, 2022, 10 a.m.