Dear hyphy developers,

I noticed that when bootstraping to test if Fst values are different from zero in compartmentilization analysis, the newest version of HYPHY (0.9920060327) forces Fst values to be positive ???.

In contrast, previous version (0.9920050429) allowed Fst values to be negative (which sometimes indeed happen in real life due to sampling effect, even though theoretically Fst can not less than zero).

As a result, p-values for the same Fst value obtained from the two different HYPHY versions are very different (the old version allowing for negative Fst seems to give more plausible results; as well as much nicer curves of cumulative probability ::)).

Is there a reason to force bootstrapped Fst values to be greater than zero  ???

regards,

Jon

Dear Jon,

There haven't been any modifications to the F_st analysis since last October - I am a bit surprised that you are getting different results. It could have something to do with the way distances (and especially ambiguities) are counted.

Could you e-mail me (help AT datamonkey org) the file which is giving odd results and the listing of old results (if you have them) - I'd like to trouble shoot this.

Thanks,
Sergei

Dear Jon,

There was a bug in the way new F_st was permuting sequence groups (note that F_st values themselves were being computed correctly). This bug was fixed in the 0.9920060524 build. Thanks so much for pointing out the issue.

Cheers,
Sergei