Swofford-Olsen-Waddell-Hillis Test

This function computes the Swofford–Olsen–Waddell–Hillis (SOWH) test Swofford, Olsen, Waddell, and Hillis (1996); Goldman, Anderson, and Rodrigo (2000) , a parametric bootstrap test. The function is computational very demanding and likely to be very slow.

Usage

SOWH.test(x, n = 100, restricted = list(optNni = FALSE), optNni = TRUE,
  trace = !getOption("quiet"), ...)

Arguments

x: an object of class "pml".
n: the number of bootstrap replicates.
restricted: list of restricted parameter settings.
optNni: Logical value indicating whether topology gets optimized (NNI).
trace: Show output during computations.
...: Further arguments passed to "optim.pml".

Value

an object of class SOWH. That is a list with three elements, one is a matrix containing for each bootstrap replicate the (log-) likelihood of the restricted and unrestricted estimate and two pml objects of the restricted and unrestricted model.

Details

SOWH.test performs a parametric bootstrap test to compare two trees. It makes extensive use simSeq and optim.pml and can take quite long.

References

Goldman N, Anderson JP, Rodrigo AG (2000). “Likelihood-based tests of topologies in phylogenetics.” Systematic biology, 49(4), 652–670.

Swofford D, Olsen G, Waddell P, Hillis D (1996). “Phylogenetic Inference.” In Hillis D, Moritz C, Mable B (eds.), Molecular Systematics, Second edition. Sinauer, Sunderland, MA.

Author

Klaus Schliep klaus.schliep@gmail.com

Examples


# in real analysis use larger n, e.g. 500 preferably more
if (FALSE) { # \dontrun{
data(Laurasiatherian)
dm <- dist.logDet(Laurasiatherian)
tree <- NJ(dm)
fit <- pml(tree, Laurasiatherian)
fit <- optim.pml(fit, TRUE)
set.seed(6)
tree <- rNNI(fit$tree, 1)
fit <- update(fit, tree = tree)
(res <- SOWH.test(fit, n=100))
summary(res)
} # }