Page 1 of 1

[astro-ph/0509478] Non-parametric reconstruction of the prim

Posted: September 19 2005
by Antony Lewis
This paper attempts to reconstruct the large scale primordial power spectrum from WMAP using a smoothing prior.

The results look a bit odd to me. Are they doing the right simulations? i.e. from Section 3 it sounds as though they are simulating posterior C_l from Eq. 18, whereas in the frequentist analysis I'd have though they should be simulating skies, computing psuedo-C_l estimators, then using these for the mock Monte Carlo reconstructions.

This might explain why they seem to get biased results when using the correct theory-model scaling in the posterior, otherwise this is also rather odd.

Since the reconstruction is likely very sensitive to the quadrupole likelihood function, would be nice if they tried using the more conservative code of Ref. astro-ph/0404567 (as used in e.g. the SDSS and B03 analyses).