Apologies if this has already been discussed but I've seen some spurious model likelihoods from the WMAP code, specifically beyond about 2sigma there are some parameter combinations that give a likelihood in excess of 200 ln units better than the maximum?
e.g. WMAP3 (TT,TE,EE) data only:
lhood \Omega_b h^2 \Omega_c h^2 \theta \tau n_s ln A_s
5373.072 0.016048 0.177486 1.056867 0.501029 1.078956 3.022956
5410.972 0.0161 0.156035 1.016962 0.462042 1.050303 2.878407
the best fit ln like is about 5626.
Spurious points in WMAP3 likelihood code

 Posts: 2
 Joined: August 02 2006
 Affiliation: Cavendish Astrophysics, Cambridge

 Posts: 23
 Joined: February 21 2006
 Affiliation: McGill University
Spurious points in WMAP3 likelihood code
Hi,
I have checked your models; they give large, negative values of TT beam and point source corrections, which almost cancel the likelihoods from TT spectrum. This may be due to the failure of the gaussian approximations in the TT beam and point source corrections. See the topic "WMAP likelihood approximations," where I gave a model with ln L = 5302 (k_{pivot}=0.002).
Loison Hoi
2 Aug 2006
I have checked your models; they give large, negative values of TT beam and point source corrections, which almost cancel the likelihoods from TT spectrum. This may be due to the failure of the gaussian approximations in the TT beam and point source corrections. See the topic "WMAP likelihood approximations," where I gave a model with ln L = 5302 (k_{pivot}=0.002).
Loison Hoi
2 Aug 2006