CosmoMC does not reproduce CAMB results

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Matteo Puel
Posts: 2
Joined: October 16 2019
Affiliation: McGill University

CosmoMC does not reproduce CAMB results

Post by Matteo Puel » July 07 2022

Hello everyone,

I am using CosmoMC (version October 2019) with a modified version of CAMB (base version = 1.0.9) where two additional parameters are included, along with two set of linear perturbation hierarchies. I modified accordingly CosmoMC to scan over such new parameters as well as the LCDM ones.

For some values of these new parameters and setting the others to the Planck 2018 values, the inferred CMB power spectra and matter power spectrum should match the LCDM expectations by physical reasons.

If I run my modified version of CAMB with one of such combination of these new parameters, the CMB power spectra and the matter power spectrum I got match perfectly the LCDM results as expected, without any error from CAMB.
If I run instead CosmoMC (which includes my modified version of CAMB) with action = 4 to test the likelihood of a model with the same choice of parameters above, I get that the model is rejected with likelihood = 0 and with a CAMB error associated to the "Dverk" integration routine.

I do not understand why there is such a difference because CosmoMC should take exactly the same CAMB version present in CosmoMC/camb directory and run CAMB_GetResults(). Since my modified CAMB seems to work fine, I am wondering why CosmoMC has having issues with the same combination of parameters.

In compiling CosmoMC, I "make" first in CosmoMC/camb/fortran directory and then in CosmoMC/source directory.

Does someone have an idea what could be wrong with CosmoMC?
I checked that the input parameters of CosmoMC (those present in .ini files) match those I use in running CAMB alone.

Thank you very much in advance for your help.

Best regards,
Matteo

Antony Lewis
Posts: 1936
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: CosmoMC does not reproduce CAMB results

Post by Antony Lewis » July 07 2022

I would check camb is actually being passed the parameters you think it is being passed

Matteo Puel
Posts: 2
Joined: October 16 2019
Affiliation: McGill University

Re: CosmoMC does not reproduce CAMB results

Post by Matteo Puel » July 07 2022

Hi Antony,

Thank you very much for your quick reply.

I already checked it by printing the values of the input parameters inside the CAMB_GetResults() function in fortran/camb.f90 file, just before the subroutine "cabmain" is called. When I run CosmoMC with action = 4 I see the same parameters that are passed to CAMB when the latter is run alone.

Do you have other suggestions or possible checks to do?

Thanks again for your help.

Best regards,
Matteo

Post Reply