Fit a genealogical placement GMRF model using maximum likelihood

Fits the COD GMRF model using the `mgcv::gam` method. Additional arguments can be passed to `gam`; see documentation for that method. Using method="REML" can speed execution by using a constrained maximum likelihood approach. Additionally, an approximate reduced-rank MRF model can be fitted by supplying the `k` parameter. If tau is not provided, `codls` is also used to optimise this parameter. This method is slower than `codls` and is not recommended for trees with more than several hundred samples.

Usage

codbinomial(tr1, logtau = NULL, k = Inf, profcontrol = list(), ...)

Arguments

tr1

Phylogenetic tree in ape::phylo format

logtau

Precision parameter. If NULL, will invoke `tauprofile` to find best value.

k

Fits a reduced-rank MRF model if k is an integer < number of nodes in the input tree. This can speed calculation but reduces precision.

profcontrol

Optional list of arguments passed to `tauprofile`

...

Additional arguments are passed to `mgcv::gam`

Value

A COD GMRF model fit. Includes the GAM model fit.

Details

The ML COD GMRF method does not currently support inverse probability weighting of samples. Use `codls` if sample weighting is needed.