Page 1 of 1

sampling method for PlanckTP + Planck SZ(SZ_plus_CMB.ini)

Posted: September 21 2017
by Zhen Pan
I am testing

Code: Select all

sampling_method = 7
and run
mpirun -np 6 ./cosmomc xx.ini
but it seems to be a bad choice: the chains only move 1-3 steps in the first 30 mins.

Anyone knows which sampling method to use here ?

Re: sampling method for PlanckTP + Planck SZ(SZ_plus_CMB.ini

Posted: September 21 2017
by Antony Lewis
You can use method 1 to use standard Metropolis Hastings.

Also check your parameter propose widths: make sure they are small enough so that chain can move. (the SZ likelihood may also be slow, I'm not sure, I've not written or used it myself).

sampling method for PlanckTP + Planck SZ(SZ_plus_CMB.ini)

Posted: September 21 2017
by Zhen Pan
Thank you, Antony. I am using method=1 now.