Page 1 of 1

Tables generated by GetDist from MCMC chains produced with Cobaya code

Posted: September 01 2020
by Claudia Scoccola
Dear all,

I have a question related to the information displayed in the tables generated by GetDist. I am analysing with GetDist some chains produced using Cobaya. I would like to know what the [math] value shown inside the parenthesis, besides the chi2 value, represents.

For illustration, I add below a piece of code. Thanks in advance for your replies.

Regards,

Claudia


info = yaml_load_file(yaml_file)

from getdist.mcsamples import MCSamples

#Load samples from chain:
gd_sample = loadMCSamples(os.path.abspath(info["output"]), settings={'ignore_rows':0.3})

gd_sample.getTable().tableTex()

Re: Tables generated by GetDist from MCMC chains produced with Cobaya code

Posted: September 01 2020
by Antony Lewis
It's a measure of the effective degrees of freedom, defined by sigma^2/2 where sigma^2 is the variance of the chi2 value.

Re: Tables generated by GetDist from MCMC chains produced with Cobaya code

Posted: September 01 2020
by Claudia Scoccola
Thank you, Antony,

I have noticed that this number ([math]) changes in different runs, even if I repeat the MCMC with the same setting (priors, reference PDF, proposal...) Also it changes if I change the reference PDF (despite I run 8 chains, and the covariance matrix of the parameters is updated on the flight).

Is there a way to use this ([math]) to decide if the posterior PDFs for the parameters are fine? Should it be close to the chi2 value? What can I conclude if this is larger or smaller than the chi2 value?

Thanks again for your comments.

Claudia