Dear Antony and all,
Suppose I want to draw samples for a set of parameters which obey a given multivariate Gaussian distribution, is it easy to hack CosmoMC to do this? I see there is a patch for this purpose for early cosmomc versions (https://sites.google.com/site/adambmant ... rs#version), and I am wondering if it is easier to do it with new versions? It would be great if you could help with instructions.
Many thanks,
GB
CosmoMC: drawing samples according to a MVNormal distribution
-
- Posts: 73
- Joined: January 04 2005
- Affiliation: NAOC
- Contact:
-
- Posts: 1944
- Joined: September 23 2004
- Affiliation: University of Sussex
- Contact:
Re: CosmoMC: drawing samples according to a MVNormal distrib
You can set 'test_likelihood=T. This will use as a likelihood your prior bounds and a Gaussian given by the covariance matrix (propose matrix) supplied.
Or you can use generic_mcmc=T and write your own in Generic_GetLogLikeMain.
Or you can use generic_mcmc=T and write your own in Generic_GetLogLikeMain.