Joined: 21 Feb 2006
Affiliation: McGill University
|Posted: October 31 2006
|Has anyone tried to comapre the values of ln(L) of the new WMAP3 code (version 2.2.1) with previous versions?
Here are the best fit likelihoods I got from version 2.1 WMAP3 code (May 2006 CosmoMC):
If I evaluate the above best fit models with the new code (version 2.2.1, Oct 2006 CosmoMC), I will have
The evidence for the running model becomes smaller, but this is not my question. The WMAP_3yr_options.f90 says
! If you'd like to compare the values of ln(L) returned by this version
! of the code with previous versions, use the following:
!! double precision, parameter :: tt_pixlike_lndet_offset(3:4) &
!! = (/5476.3672001139175d0,29467.570953238155d0/)
!! double precision, parameter :: teeebb_pixlike_lndet_offset = 0d0
!! double precision, parameter :: te_lndet_offset = 0d0
I used the above settings and got
The Δ χ2 are almost the same, as I would have expected. However,
1. The values of χ2 are negative, is it a problem?
2. I would expect that I will have a number like -log(like)=5626 so that the new value of likelihood is comparable to the old value. However, I got a number like −18057 instead. I also tried to use
!! integer :: lowl_tt_res = 3
!! integer :: lowl_max = 12
And the numbers are -log(like)=1668.8 for the default offset settings, and -log(like)=4580.6 for the above offset settings. So, does anyone know why I did not have the numbers which are comparable to the old likelihoods? Thanks.
31 Oct 2006