Page 1 of 1

[CosmoMC/Planck likelihood] Precision

Posted: November 20 2018
by Luke Hart
Dear all,

Does anyone know how precise the Planck likelihood values that we get out of CosmoMC are? When looking at analysis with the likelihood values, it's really important to know what the actual precision is so we don't get dominated by noise in the numerical analysis. Does anyone know how accurate the likelihood values are (i.e. of the 6466.xxxxxx for the best fit value I have for a given parameter set, what is the number of d.p. one can go to before this becomes noise dominated)?

Thanks in advance for your help

Luke

Re: [CosmoMC/Planck likelihood] Precision

Posted: November 21 2018
by Antony Lewis
The noise is the likelihood itself is small. However there will be numerical noise in the parameters of the best-fit point, which can be O(1).

Re: [CosmoMC/Planck likelihood] Precision

Posted: December 05 2018
by Luke Hart
Thanks Antony, but I just wanted to ask: what do you mean when you say 'O(1)' ??

Re: [CosmoMC/Planck likelihood] Precision

Posted: December 06 2018
by Antony Lewis
I mean chi^2 can move around by O(1) depending on where exactly the minimizer ends up (though usually stable to somewhat below \pm 1).