Open jboyko opened 4 years ago
I think the best option is to give a warning that it's un-estimateable and add an option to get getRateMat4Dat to include dual transitions. Will add this soon.
If the MK model cannot visit some states, then you might not have an ergotic Markov Chain ( https://brilliant.org/wiki/ergodic-markov-chains/ ). I don't know if some of the assumptions of the MK model will be broken if this is the case.
Bingo! That's exactly the issue. Now, how do we make an efficient test of ergodicity?
These are probably estimable, but thrown an error during the ancestral state reconstruction. We should warn the user when this is the case and perhaps offer a default matrix where dual transitions are allowed. In my opinion, this is partially a matter of whether allowing these types transitions break the biological assumptions of the model.