CosmoCoffee

Anze Slosar · Posted: December 16 2009

This is something I always dreamt of doing... It is quite neat and the "detection" they have points at how difficult is to do this in practice... :)

My comments are:

a) Basically, it is quite clear that absolute photometric calibration is a big issue: that is probably the reason why the mean number density varies from the south to the north, etc. I don't think it is LSS contamination they are seeing, is just the non-uniform depth of the survey. Note that a single wrong patch spils over all ls in fourier space... It is also something that is most difficult to measure and correct for.

b) star-galaxy separation: there is this standard trick of marginalising over templates that has been used in CMB for ages and should work here as well: just add template covariance to you cov matrix with some large prefactor... Works like magic!

c) They claim that going to large zs helps because growth factor is lower and because galaxies are less evolving: but isn't them main effect simply from the fact that one is looking at larger scales and thus more uniform universe?

d) I also don't quite understand the hocus-pocus around contribution to diagonal elements from structure: they do some smoothing game. But instead, the eq. 16 with w

g

(0) should give the right answer... Of course, you don't know what non-linear structure is doing exactly, but you can always parametrise this sensibly and then marginalise over it. Data should be able to constrain this internally (if all you want is dipole), I think.

I also think that LSST should be much better for this, not only because it has full sky coverage, but also, because of its scanning strategy it will revisit each patch of the sky many times, which should give much more uniform depth by simple averaging over atmospheric conditions...