Minor weirdness in inferring conflicted variants between focal sites

[hyanwong]

In

tsinfer/tsinfer/algorithm.py

Line 201 in 62b0135

if ones >= zeros:

in the case of equal numbers of derived and ancestral types at an older site (which can only happen if there is a conflict), we take the derived state. I reckon this should probably be ones > zeros instead. It's possible that we could also do something more sophisticated for ancient samples by taking into account the sample age.

[hyanwong]

After discussion with @jeromekelleher it seems like we previously experimented with both ones > zeros and ones >= zeros, and decided on the >= version, so closing this and leaving as-is.