Current pipeline estimates divergence times using R packages before input into model. Methods are questionable, chronos() takes too long, chronoMPL() produces negative branch lengths, and force.ultrametric() is rough estimate.
Instead, model branch lengths as parameters in stan model. Starting at root, branches are proportions of total (lower=0,upper=1), with no priors. Branch lengths based on sequence divergence can be modeled as data based on the the temporal parameters. May or may not allow for variation in rates.
rmcminds
commented
3 years ago
Current pipeline estimates divergence times using R packages before input into model. Methods are questionable, chronos() takes too long, chronoMPL() produces negative branch lengths, and force.ultrametric() is rough estimate.
Instead, model branch lengths as parameters in stan model. Starting at root, branches are proportions of total (lower=0,upper=1), with no priors. Branch lengths based on sequence divergence can be modeled as data based on the the temporal parameters. May or may not allow for variation in rates.