cosmomc : getdist marginalised probability

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
Vimal Simha
Posts: 13
Joined: May 11 2006
Affiliation: Ohio State University

cosmomc : getdist marginalised probability

Post by Vimal Simha » October 24 2006

Hi,

I was under the impression that:

a) filename.margestats showed statistics from the marginalised posterior pdf.

b) the solid line plotted from filename_px.dat plotted the marginalised posterior pdf.

However, I am finding that there is a conflict between these. The value of param for which column 2 (which is the normalised likelihood) of *.dat =1 is not the mean value in file.margestats

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: cosmomc : getdist marginalised probability

Post by Antony Lewis » October 24 2006

The maximum of the posterior is not the same as the mean unless the posterior happens to be exactly symmetric.

Vimal Simha
Posts: 13
Joined: May 11 2006
Affiliation: Ohio State University

cosmomc : getdist marginalised probability

Post by Vimal Simha » October 27 2006

Thanks for that clarification.

1) I understand that the limits produced by margestats are based on having an equal number of samples on either side.

2) Is it possible to get limits based on the area enclosed at a particular likelihood from getdist or would we have to compute it from the chain(s) ourselves ?

Antony Lewis
Posts: 1941
Joined: September 23 2004
Affiliation: University of Sussex
Contact:

Re: cosmomc : getdist marginalised probability

Post by Antony Lewis » October 27 2006

There's no option to calculate the other choice of limit definition at the moment.

Post Reply