phangorn (version 1.5-0)

pmlCluster: Stochastic Partitioning

Description

Stochastic Partitioning of genes into p cluster.

Usage

pmlCluster(formula, fit, weight, p=1:5, part=NULL, nrep = 10, control=pml.control(epsilon=1e-8, maxit=10, trace=1),...)

Arguments

formula

a formula object (see details).

fit

an object of class pml.

weight

weight is matrix of frequency of site patterns for all genes.

p

number of clusters.

part

starting partition, otherwise a random partition is generated.

nrep

number of replicates for each p.

control

A list of parameters for controlling the fitting process.

...

Further arguments passed to or from other methods.

Value

• pmlCluster returns a list with elements
• logLik
  log-likelihood of the fit
• trees
  a list of all trees during the optimization.
• fits
  fits for the final partitions

Details

The formula object allows to specify which parameter get optimized. The formula is generally of the form edge + bf + Q ~ rate + shape + ..., on the left side are the parameters which get optimized over all cluster, on the right the parameter which are optimized specific to each cluster. The parameters available are "nni", "bf", "Q", "inv", "shape", "edge", "rate". Each parameter can be used only once in the formula. There are also some restriction on the combinations how parameters can get used. "rate" is only available for the right side. When "rate" is specified on the left hand side "edge" has to be specified (on either side), if "rate" is specified on the right hand side it follows directly that edge is too.

See Also

pml,pmlPart,pmlMix,SH.test

Examples

data(yeast)
dm <- dist.logDet(yeast)
tree <- NJ(dm)
fit=pml(tree,yeast)
fit = optim.pml(fit)

weight=xtabs(~ index+genes,attr(yeast, "index"))
set.seed(1)

sp <- pmlCluster(edge~rate, fit, weight, p=1:4)
sp
SH.test(sp)