Optimise smoothing parameter for a given tree and branch statistic

This function maximises the deviance explained of future evolution of the given branch statistic via its dependence on a smoothing parameter (tau). This function must be given a tree and a function (e.g. `codls` or `lbi`) which takes a tree and a smoothing parameter and returns a given branch statistic. The association of the branch statistic and future coalescent events (prediction of evolution) is modelled using a GAM and adjusting for branch lengths and distance from the root.

Usage

optsmooth(
  tr,
  func,
  logtaulb = -4,
  logtauub = 35,
  startpc = 50,
  endpc = 100,
  nobj = 100
)

Arguments

tr: A phylogenetic tree in ape::phylo format
logtaulb: Lower bound of precision parameteters
logtauub: Upper bound of precision parameteters
startpc: The initial per cent of nodes in the tree counting from root to tips where the loss function will be evaluated
endpc: The final per cent of nodes in the tree counting from root to tips where the loss function will be evaluated
nobj: The integer number of points along the tree where the loss function will be evaluated. If Inf, will use all points between startpc and endpc, but may be slow.

Value

Output of `optimise`. $minimum contains the optimal smoothing parameter