CosmoMC: Fast/slow parameters

Use of Cobaya. camb, CLASS, cosmomc, compilers, etc.
Post Reply
gongbo zhao
Posts: 73
Joined: January 04 2005
Affiliation: NAOC
Contact:

CosmoMC: Fast/slow parameters

Post by gongbo zhao » January 19 2015

Dear Antony and all,

I am having a question for the sampling of fast/slow parameters. I have many fast parameters (>40). When I sample them together with the usual slow parameters (\Omega_b h^2, \Omega_c h^2, \theta, etc), they are sampled well but the slow parameters are sort of stuck in the parameter space. Any idea why this happens and is there an easy fix?

Btw, I am using the default fast/slow sampling setting as below,

sampling_method = 7
dragging_steps = 3
propose_scale = 2

Thanks,
GB

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

Re: CosmoMC: Fast/slow parameters

Post by Antony Lewis » January 19 2015

Not waited for it to converge, or the slow parameter initial steps sizes/covariance are wrong? Do you have lots of out put samples?

You can decrease dragging_steps to balance things more in favour of the slow parameters.

gongbo zhao
Posts: 73
Joined: January 04 2005
Affiliation: NAOC
Contact:

CosmoMC: Fast/slow parameters

Post by gongbo zhao » January 19 2015

Thanks Antony. The covariance is correct for the slow parameters and I have normal output samples. I will experiment on dragging_steps.

Thanks again,
GB

Post Reply