Given a bammdata object and a vector of (continuous)
trait data, assess whether the correlation between the trait and bamm
estimated speciation, extinction or net diversification rate is
significant using permutation. A set of posterior samples is randomly
drawn from the bammdata object. If the trait is continuous,
this function calculates the correlation coefficients between the
trait and tip rates (observed correlation), as well as that with
permuted rates for each posterior sample. In a one-tailed test for
positive correlations, the reported p-value is the proportion of the
posterior samples in which the observed correlation is larger than the
correlations calculated with permuted rates. In a two-tailed test, the
reported p-value is the proportion of the posterior samples in which
the null correlation is as extreme as the correlation observed. If the
trait is binary, the U statistic of the Mann-Whitney test is
calculated instead of correlation coefficients to assess whether there
is a significant difference in rate between the two trait states. For
categorical traits with more than two states, the Kruskal-Wallis rank
sum statistic is used.
traitDependentBAMM(
ephy,
traits,
reps,
rate = "speciation",
return.full = FALSE,
method = "spearman",
logrates = TRUE,
two.tailed = TRUE,
traitorder = NA,
nthreads = 1
)A list with the following components:
estimate: A numeric value for continous trait data: the average observed correlation between tip rates and the trait across the posterior samples. For categorical traits, it is a list showing the median species-specific rates for each trait state.
p.value: A numeric value. The probability that the observed correlation is less than or equal to a sample from the null distribution.
method: A character string, as input.
rate: A character string, as input.
two.tailed: A logical, as input.
gen: An integer vector, recording which posterior samples
were selected. Only present when return.full is
TRUE.
obs.corr: A numeric vector, the observed correlation
coefficents for each posterior sample. Only present when
return.full is TRUE. For binary traits, centered
U statistics (U - n1* n2/2; where n1 and n2 are the number of
species in each state of the binary trait) is reported.
null: A numeric vector. The null distribution of
correlation coefficients (or centered U statistics for binary
traits) from permutation. Only present when return.full
is TRUE.
An object of class bammdata.
A vector of trait data, with names corresponding to tips in
the bammdata object. It can be numeric or categorical.
An integer specifying the number of permutations (i.e., the
number of posterior samples to randomly draw with replacement from the
bammdata object).
A character string specifying which estimated rate from the
bammdata object to use for testing correlation, must be one of
'speciation', 'extinction', 'net diversification' or 'trait'. Defaults to
'speciation'. You can specify just the initial letter. Ignored for
trait event data.
A logical. If TRUE, the list of posterior
samples, the observed correlation for each posterior sample, and the
null distribution will be included in the returned object. Defaults to
FALSE.
A character string, must be one of 'spearman', 'pearson', 'mann-whitney', or 'kruskal'. Defaults to 'spearman'. You can specify just the initial letter.
A logical. If TRUE log-transform the rates before
analysis. Defaults to TRUE. This can only matter for the
pearson correlation.
A logical, used for continuous trait data. If
TRUE, perform a two-tailed statistical test (i.e., if the null
distribution is symmetric, it is equivalent to doubling the p-value).
Defaults to TRUE.
A character string specifying the direction of correlation for the alternative hypothesis. For continuous traits, it must be either "positive" or "negative"; only the initial letter is needed. For binary traits, it must be a string indicating states with increasing rate under the alternative hypothesis, separated by comma (e.g., 'A, B'). One-tailed test for categorical data with more than two states is not supported.
Number of threads to use for parallelization of the
function. The R package parallel must be loaded for
nthreads > 1.
Dan Rabosky, Huateng Huang
Tip rates --trait, speciation, extinction, or net diversification
rates-- are permuted in a way such that pairwise covariances in rates
between species are maintained. That is, tips with the same
tipStates still have the same rate after permutation. Posterior
samples are randomly selected with replacement from the
bammdata object, so reps could be smaller or larger than
the total number of samples in the object.
This function expects that the bamm-data object and the trait data
have the same taxon set. It may be necessary to subset the trait data
and/or run subtreeBAMM on the bamm-data object in
order to meet this requirement.
Rabosky, D. L. and Huang, H., 2015. A Robust Semi-Parametric Test for Detecting Trait-Dependent Diversification. Systematic Biology 65: 181-193.
Rabosky, D. L. 2014. Automatic detection of key innovations, rate shifts, and diversity-dependence on phylogenetic trees. PLoS ONE 9:e89543.
Rabosky, D. L., F. Santini, J. T. Eastman, S. A. Smith, B. L. Sidlauskas, J. Chang, and M. E. Alfaro. 2013. Rates of speciation and morphological evolution are correlated across the largest vertebrate radiation. Nature Communications DOI: 10.1038/ncomms2958.
subtreeBAMM
# using a small subset of the fish data set (300 taxa) in Rabosky et al.
# 2013. Nat. Com. paper
data(fishes, events.fishes)
xx <- getEventData(phy = fishes, eventdata = events.fishes,
nsamples = 500, type = "diversification")
# traits.fishes is the trait -- body size
data(traits.fishes)
x <- traitDependentBAMM(ephy = xx, traits = traits.fishes, reps = 1000,
return.full = TRUE, method = 's', logrates = TRUE,
two.tailed = TRUE)
Run the code above in your browser using DataLab