Fast (parallel) POUMM likelihood calculation using the SPLITT library

Fast (log-)likelihood calculation using C++ and OpenMP based parallelization.

likPOUMMGivenTreeVTipsC(
  integrator,
  alpha,
  theta,
  sigma,
  sigmae,
  g0 = NA,
  g0Prior = NULL,
  log = TRUE
)

Arguments

integrator	An Rcpp module object from the class POUMM_AbcPOUMM. This object is to be created using the function pruneTree (see example). This object contains the data and tree (arguments z and tree of the equivalent function dVTipsGivenTreeVTips.).
alpha	the strength of the selection
theta	long term mean value of the OU process
sigma	the unit-time standard deviation of the random component in the OU process.
sigmae	the standard deviation of the environmental deviation added to the genetic contribution at each tip, by default 0, meaning no environmental deviation.
g0	Numeric, NA or NaN, either a fixed genotypic value at the root of tree or NA or NaN. A NA "Not Available" will cause to analytically calculate the value of g0 that would maximize the conditional likelihood of the data given g0. A NaN "Not a Number" will cause integration over g0 taking values in (-Inf,+Inf) assuming that g0 is normally distributed with mean g0Prior$mean and variance g0Prior$var (see parameter g0Prior).
g0Prior	Either NULL or a list with named numeric or character members "mean" and "var". Specifies a prior normal distribution for the parameter g0. If characters, the members mean and var are evaluated as R-expressions.
log	Logical indicating whether log-likelihood should be returned instead of likelihood, default is TRUE.

Value

A numeric with attributes "g0" and "g0LogPrior".

Details

This function is the C++ equivalent of dVTipsGivenTreePOUMM (aliased also as likPOUMMGivenTreeVTips). Currently, the function does not support multiple precision floating point operations (supported in dVTipsGivenTreePOUMM). The C++ implementation is based on the library for parallel tree traversal "SPLITT" (https://github.com/venelin/SPLITT.git).

References

Mitov, V., and Stadler, T. (2017). Fast and Robust Inference of Phylogenetic Ornstein-Uhlenbeck Models Using Parallel Likelihood Calculation. bioRxiv, 115089. https://doi.org/10.1101/115089 Mitov, V., & Stadler, T. (2017). Fast Bayesian Inference of Phylogenetic Models Using Parallel Likelihood Calculation and Adaptive Metropolis Sampling. Systematic Biology, 235739. http://doi.org/10.1101/235739

Examples

if (FALSE) {
N <- 100
tr <- ape::rtree(N)
z <- rVNodesGivenTreePOUMM(tr, 0, 2, 3, 1, 1)[1:N]
pruneInfo <- pruneTree(tr, z)
microbenchmark::microbenchmark(
  likCpp <- likPOUMMGivenTreeVTipsC(pruneInfo$integrator, 2, 3, 1, 1),
  likR <- likPOUMMGivenTreeVTips(z, tr, 2, 3, 1, 1, pruneInfo = pruneInfo))

# should be the same values
likCpp
likR
}