move PhoSin QC where it should go

[petrelharp]

There was a whoopsie in #1560 I didn't notice - the code was in qc/PhoSin.py not stdpopsim/qc/PhoSin.py. Fixing this.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 99.83%. Comparing base (1213dc1) to head (c1d44ca).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #1561   +/-   ##
=======================================
  Coverage   99.83%   99.83%           
=======================================
  Files         131      131           
  Lines        4334     4340    +6     
  Branches      595      595           
=======================================
+ Hits         4327     4333    +6     
  Misses          3        3           
  Partials        4        4           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[petrelharp]

Looks like the parameters for the gamma disagree:

>           assert np.allclose(mt1[i].distribution_args, mt2[i].distribution_args)
E           AssertionError: assert False
E            +  where False = <function allclose at 0x7fb9680175e0>([-0.00971, 0.131], [-0.0257, 0.131])
E            +    where <function allclose at 0x7fb9680175e0> = np.allclose
E            +    and   [-0.00971, 0.131] = MutationType(dominance_coeff=None, distribution_type='g', distribution_args=[-0.00971, 0.131], convert_to_substitution=True, dominance_coeff_list=[0.0, 0.01, 0.1, 0.4], dominance_coeff_breaks=[-0.1, -0.01, -0.001]).distribution_args
E            +    and   [-0.0257, 0.131] = MutationType(dominance_coeff=None, distribution_type='g', distribution_args=[-0.0257, 0.131], convert_to_substitution=True, dominance_coeff_list=[0.0, 0.01, 0.1, 0.4], dominance_coeff_breaks=[-0.1, -0.01, -0.001]).distribution_args

This issue was flagged by @igronau in #1548:

The gamma shape was taken directly from the reported estimate.
The gamma mean was taken from the value reported in simulations. 
This was derived from the inferred shape and scale + the estimated Na, 
but since the estimated Na is not reported, I couldn't confirm the reported mean.

@ckyriazis, can you confirm/clarify?

[ckyriazis]

ah oops sorry about the directory issue! yes, this is what I was expecting - the gamma mean in the original PR was incorrect, -0.0257 is the correct value.

[petrelharp]

Hm, I do see that in the paper:

Selection coefficients (s) for deleterious mutations were drawn from a gamma
distribution estimated using the genomic data (see Inference of the distribution of selection
coefficients) with mean = -0.0257 and shape parameter = 0.1314. 

Thanks! I will fix this and merge; if @igronau disagrees, we can discuss and fix it up separately.

[petrelharp]

Ah, it looks like it was a case of just grabbing the wrong number - the number previously in the code was from the lower mutation rate:

For sensitivity analyses
varying the mutation rate, we used the parameters inferred under the respective mutation rate
(mean = -0.00971 and shape = 0.1316 for μ = 2.2x10-9 and mean = -0.0476 and shape = 0.1315
for μ = 1.08x10-8; fig. S13).

[igronau]

Yes, this is correct. Thanks, @petrelharp for fixing this.

[igronau]

Looks good.