Choice of outgroup about ncd-statistics

[bbitarello]

Hello! The outgroup is used in NCD2 (but not NCD2). NCD2 outperforms NCD1, so is preferable. The way this information is used is as follows:

1. you only need one individual from the sister species. So if that is available, it is enough. The information you need is the number of fixed differences between that species and your species for every window you query.
2. Regarding the fact that the sister species might potentially also be under balancing selection at this locus, I have not explored this. My intuition is that then both species should have a reduced number of FDs relative to the number of SNPs. This may or may not be a problem, depending on what you are comparing your candidate locus to. If, as you said, there might be another sister group to use, then maybe do that.
3. Divergence time. You want the two species to be sufficiently diverged for you to have the signature. At least 4xNe generations, where Ne is the effective population size of your species and the sister species. For example, in humans the long-term Ne is roughly 10,000, so 4x10,000=40,000 generations. To convert this to years you need to multiply this by the generation time. If we consider this to be 25 years for humans, we get 1,000,000 (1 million years). The sister species we use for humans is chimp and human and chimp diverged from their common ancestor at least 6 million years ago, so this is appropriate.

I hope this helps!