rayDISC: Evolution of categorical traits

In corHMM: Analysis of binary character evolution

Description

Fits a model of evolution for categorical traits, allowing for multi-state characters, polymorphisms, missing data, and incompletely resolved trees

Usage

rayDISC(phy,data, ntraits=1, charnum=1, rate.mat=NULL, model=c("ER","SYM","ARD"),
node.states=c("joint", "marginal", "scaled"), p=NULL, root.p=NULL, ip=NULL, 
lb=0, ub=100, diagn=FALSE)

Arguments

phy	a phylogenetic tree, in ape “phylo” format.
data	a data matrix containing species information (see Details).
ntraits	specifies the number of traits to included in the analysis.
charnum	specified the character to analyze.
rate.mat	a user-supplied rate matrix index of parameters to be optimized.
model	specifies the underlying model.
node.states	method used to calculate ancestral states at internal nodes.
p	a vector of transition rates. Allows the user to calculate the likelihood given a specified set of parameter values to specified as fixed and calculate the likelihood.
root.p	a vector used to fix the probabilities at the root, but “maddfitz” can also be supplied to use the method of Maddison et al (2007) and FitzJohn et al (2009) (see details).
ip	initial values used for the likelihood search. Can be a single value or a vector of unique values for each parameter. The default is ip=1.
lb	lower bound for the likelihood search. The default is lb=0.
ub	upper bound for the likelihood search. The default is ub=100.
diagn	logical indicating whether diagnostic tests should be performed. The default is FALSE.

Details

The function takes a tree and a trait file and estimates transition rates and ancestral states for binary or multistate characters. The first column of the trait file must contain the species labels to match to the tree, with the second, third, fourth, and so on, corresponding to the traits of interest. Use the charnum variable to select the trait for analysis. Also, the input phylogeny need not be bifurcating as the algorithm is implemented to handle multifucations. Polytomies are allowed by generalizing Felsenstein's (1981) pruning algorithm to be the product of the probability of observing the tip states of n descendant nodes, rather than two, as in the completely bifurcating case.

The user can fix the root state probabilities by supplying a vector to the root.p. If user supplies the flag root.p=“maddfitz” calls the same procedure described by Maddison et al (2007) and FitzJohn et al (2009) is used. Note, however, that the default root.p=NULL assumes equal weighting among all possible states.

Ambiguities (polymorphic taxa or taxa missing data) are assigned likelihoods following Felsenstein (2004, p. 255). Polymorphic taxa are coded “&” with all states observed at a tip. For example, if a trait has four states and taxonA is observed to be in state 1 and 3, the character would be coded as “1&3”. rayDISC then uses this information to assign a likelihood of 1.0 to both states. Missing data are treated as ambiguous for all states, thus all states for taxa missing data are assigned a likelihood of 1.0. For example, for a four-state character (i.e. DNA), a taxon missing data will have likelihoods of all four states equal to 1.0 [e.g. L(A)=1.0, L(C)=1.0, L(G)=1.0, L(T)=1.0].

Value

rayDISC returns an object of class rayDISC. This is a list with elements:

$loglik	the maximum negative log-likelihood.
$AIC	Akaike information criterion.
$AICc	Akaike information criterion corrected for sample size.
$ntraits	The number of traits specified.
$solution	a matrix containing the maximum likelihood estimates of the transition rates.
$solution.se	a matrix containing the approximate standard errors of the transition rates. The standard error is calculated as the square root of the diagonal of the inverse of the Hessian matrix.
$index.mat	The indices of the parameters being estimated are returned. The numbers correspond to the row in the eigvect and can useful for identifying the parameters that are causing the objective function to be at a saddlepoint.
$opts	Internal settings of the likelihood search.
$data	User-supplied dataset.
$phy	User-supplied tree.
$states	The likeliest states at each internal node.
$tip.states	NULL
$iterations	The number of iterations used by the optimization routine.
$eigval	The eigenvalues from the decomposition of the Hessian of the likelihood function. If any eigval<0 then one or more parameters were not optimized during the likelihood search.
$eigvect	The eigenvectors from the decomposition of the Hessian of the likelihood function is returned.
$bound.hit	A logical for diagnosing if rate parameters were constrained by lb or ub values during optimization.
$message.tree	A list of taxa which were listed in the data matrix, but were not present in the passed phylo object. These taxa will be excluded from the analysis. message.tree is null if all taxa in data are included in tree.
$message.data	A list of taxa which were present in the passed phylo object, but lacked data in the passed data matrix. These taxa will be coded as missing data (all states equally likely). message.data is null if all taxa in tree have entries in data matrix.

Author(s)

Jeffrey C. Oliver and Jeremy M. Beaulieu

References

Felsenstein, J. 1981. A likelihood approach to character weighting and what it tells us about parsimony and compatibility. Biological Journal of the Linnean Society 16: 183-196.

Felsenstein J. 2004. Inferring phylogenies. Sunderland MA: Sinauer Associates.

FitzJohn, R.G., W.P. Maddison, and S.P. Otto. 2009. Estimating trait-dependent speciation and extinction rates from incompletely resolved phylogenies. Systematic Biology 58:595-611.

Maddison, W.P., P.E. Midford, and S.P. Otto. 2007. Estimating a binary characters effect on speciation and extinction. Systematic Biology 56:701-710.

See Also

plotRECON

Examples

# Not run
### Example 1
# Load data
# data(rayDISC.example)

## Perform ancestral state estimation, using an asymmetric model of evolution and marginal
## reconstruction of ancestral states
# recon <- rayDISC(rayDISC.example$tree,rayDISC.example$trait,model="ARD",
# node.states="marginal")

## Plot reconstructions on tree
# plotRECON(rayDISC.example$tree,recon$states)

### Example 2
## Perform ancestral state estimation on second character, using a single-rate model of 
## evolution, marginal reconstruction of ancestral states, and setting the lower bound for
##parameter estimates to 0.01
# recon <- rayDISC(rayDISC.example$tree,rayDISC.example$trait,charnum=2,model="ER",
# node.states="marginal",lb=0.01)

### Example 3
## Perform ancestral state estimation on third character, using a single-rate model of 
## evolution and joint reconstruction of ancestral states
# recon <- rayDISC(rayDISC.example$tree,rayDISC.example$trait,charnum=3,
# model="ER",node.states="joint") 

Example output

Loading required package: ape
Loading required package: nloptr
Loading required package: GenSA

corHMM documentation built on May 2, 2019, 4:48 p.m.