1D plots generated by running getdist

Use of Healpix, camb, CLASS, cosmomc, compilers, etc.
Post Reply
Archita Bhattacharyya
Posts: 4
Joined: August 02 2017
Affiliation: Indian Statistical Institute Kolkata

1D plots generated by running getdist

Post by Archita Bhattacharyya » December 27 2019

Hello all,

I have used Getdist program supplied by CosmoMC to generate some 1D and 2D contour plots. As quoted in this tutorial http://www.uco.es/~ajcuesta/cosmomc-ugr---day-3.pdf , the .likestats file generated by running getdist provides the bestfit value and the bounds of mean likelihood distribution for each parameter, whereas, .margestats file contains the mean and marginalized bound for each parameter. Now after generating the plots I have noticed that the 1D plots match the bounds as given in .likestats file, but the 2D plots as in .margestats file.

My question is, if 1D plots are marginalized posterior distribution of each parameter, aren't they supposed to match the bounds of .margestats file? In general, I want to know how the marginalized posterior distributions are generated then? Can anyone clarify this?

Thanks in advance.


Antony Lewis
Posts: 1652
Joined: September 23 2004
Affiliation: University of Sussex

Re: 1D plots generated by running getdist

Post by Antony Lewis » December 27 2019

The .likestats file has the min-max parameter values for the N-D equal-likelihood confidence contour.

.margestats marginalized bounds definition depends on your getdist settings, but by default the recipe followed is the same as in the Planck parameter papers. Any match with the .likestats file is likely a coincidence (unless you have only one parameter or things are very Gaussian).

Post Reply