search

thej022214 / corHMM

Fits a generalized form of the covarion model that allows different transition rate classes on different portions of a phylogeny by treating rate classes as “hidden” states in a Markov process.
11 stars 13 forks source link

Missing states #4

Closed thej022214 closed 4 years ago

thej022214 commented 7 years ago

Turns out different functions have different requirements for how to score missing characters. In corHMM "NA" is considered missing, whereas in rayDISC "?" assumes it is missing.

This needs to be standardized.

jcoliver commented 7 years ago

@thej022214 Yes, standardization needs to happen. You want me on this?

jcoliver commented 7 years ago

I think the approach used in corHMM is the most appropriate, by the way (NA = missing).

thej022214 commented 7 years ago

JEFF!!! I am not sure what corDISC uses, but I can deal with that one. But if you wanted to take a crack at rayDISC that would be sweet. Note, that I've made some new changes to rayDISC, though it should not impact what you are doing.

thej022214 commented 7 years ago

I just fixed corHMM. All seems to work fine. Had to deal with issues with phangorn and generating starting values, but other than that, seems good.

jcoliver commented 7 years ago

Updated rayDISC with similar fix as corHMM, while still supporting "?" as missing state (if user has ?, it gets replaced by NA before analysis). Also made some updates to conform to variable names in corHMM.

thej022214 commented 7 years ago

Cool! I will check it out. I am sure you noticed, but I had an issue with phangorn and starting values when there was an NA. I don't remember what rayDISC does for a starting value, but I will take a look.

thej022214 commented 4 years ago

Should be fixed.