CosmoMC finding best-fit params
Posted: March 02 2017
Hi,
I was trying to use CosmoMC action=2 to find best-fit parameters for WMAP data. To make the results more reliable, I ran 4 minimizations at the same time as suggested in the CosmoMC README. I got the best-fit points without warning.
However, when I looked closer at each set of parameters given by the minimizations, although their Chi2 were very close, the parameters differed from one another quite significantly (at some fractions of a sigma). I think it's possible that the best-fit points I got were just local minimums. If so, is there some way to avoid them?
Below is the setting I used for the input file:
#when estimating best fit point (action=2 or estimate_propose_matrix),
#required relative accuracy of each parameter in units of the covariance width
max_like_radius = 0.0001
max_like_iterations = 100000
minimization_points_factor = 2
minimize_loglike_tolerance = 0.0005
minimize_separate_fast = T
#if non-zero do some low temperature MCMC steps to check minimum stable
minimize_mcmc_refine_num = 20
minimize_refine_temp = 0.01
minimize_temp_scale_factor = 5
minimize_random_start_pos = T
Is there anything I can do to make the results from action=2 more reliable?
Thanks a lot !
Yajing
I was trying to use CosmoMC action=2 to find best-fit parameters for WMAP data. To make the results more reliable, I ran 4 minimizations at the same time as suggested in the CosmoMC README. I got the best-fit points without warning.
However, when I looked closer at each set of parameters given by the minimizations, although their Chi2 were very close, the parameters differed from one another quite significantly (at some fractions of a sigma). I think it's possible that the best-fit points I got were just local minimums. If so, is there some way to avoid them?
Below is the setting I used for the input file:
#when estimating best fit point (action=2 or estimate_propose_matrix),
#required relative accuracy of each parameter in units of the covariance width
max_like_radius = 0.0001
max_like_iterations = 100000
minimization_points_factor = 2
minimize_loglike_tolerance = 0.0005
minimize_separate_fast = T
#if non-zero do some low temperature MCMC steps to check minimum stable
minimize_mcmc_refine_num = 20
minimize_refine_temp = 0.01
minimize_temp_scale_factor = 5
minimize_random_start_pos = T
Is there anything I can do to make the results from action=2 more reliable?
Thanks a lot !
Yajing