Open AudeCaizergues opened 3 hours ago
Hi Aude,
S_b
to an arbitrary value to avoid optimizing an unneeded (and unidentifiable) parameter when fixing p_b
to 0.S_b
to an arbitrary value here, and of course check if they were indeed held fixed by inspecting JointInference.params_mle
.marginal
component spectra (types "urb"
and "rur"
in your case). You can obtain the marginal inference objects by accessing JointInference.marginal_inferences
which is a dictionary of BaseInference
objects for the different types. (You can ignore the "all"
entry which is the BaseInference
results for the sum of all component spectra.) The marginal inference inherit the fixed parameters from JointInference
(which can you verify by checking JointInference.marginal_inferences[type].fixed_params
), so what you're effectively testing is whether eps
varies between rural and urban populations (which might not be so informative after all). If you want to test whether purifying selection differs, you would share S_d
or b
or both between your types.I hope this helps Janek
Hello,
I am trying to run a joint inference from the SFS of 2 different populations and I have some questions regarding the process.
From the nature of my data and my questions, I only want to estimate purifying selection occurring in my 2 populations. I understand that we can constrain parameters to do so such as p_b=0 so that S can't be positive. In the manual however (section DFE inference > Base inference > Fixing parameters), I read that "We also force the DFE to be purely deleterious, by fixing S_b to some arbitrary value and setting p_b to 0". I understand why we fix p_b to 0 but I wonder why we necessarily have to fix S_b=1 (in this case) to achieve the estimation of only deleterious mutations. Could you please elaborate on that ?
I found no example combining joint inference with fixing parameters so I code my script like that (where eps is shared and p_b fixed to 0, I'll add S_b=1 if need be):
Is it the proper way to (visually) test some difference in selection between my populations ?
When I run this test on my data I get:
But I have a hard time figuring out what it means in terms of eps and p_b and hesitate between two interpretations :
Thank you in advance for your help,
Aude