Directly handling frequencies #4
Description
Copied from FangfeiLi05#8
FitSeq currently expects counts and does a simplex normalization to frequency. I think.
However, it may be problematic to have really low counts lineages for the modeling, at least for using likelihood to assess fit.
It would be advantageous to permit just inputting the frequencies, and filter crap lineages so we don't waste time on fitting those.
Is this possible? How then is average fitness estimated?
After the first estimate, can this parameter be fit (ie not just the weighted average of the lineages?)