Closed andreufont closed 6 years ago
That's strange. Could we see a plot of the result?
First plot is from my run, using a very old simulation (from December, before the fix of mean delta).
The second plot is from @jfarr03, using a very recent simulation after merging revamp into master:
all_wedges_cf_exp_10_revamp_density_rmax160_n40_n40.pdf
The wedges are slightly different, but you can see that both are anisotropic, and look like linear RSD.
Actually, forget about it for now. There is clearly something off with these plots (they do not agree at all!), so we should spend some time figuring this out first. We'll come back to you once we understand better the plots.
I can't really see how this would be caused by CoLoRe, so my first reaction would be to blame it on post-processing or the 2pcf code. The only effect that is unique about the line of sight in CoLoRe is the fact that the density field that gets interpolated onto the skewers is in the lightcone (i.e. it evolves with r), but I doubt that this would give you such strong effects.
Have you guys tried to run the 2pcf code on the raw CoLoRe skewers without post-processing? Also, not sure whether the 2pcf code has to make assumptions about cosmology. If so, and if you pass the wrong cosmological parameters, I wonder if the AP effect could cause this (this feels like a long shot though).
Yes, I was also very surprised, since these measurements are directly on the skewer density (they didn't go through the DESI simulation code). I think we should test:
We'll come back to you!
@damonge @fjaviersanchez: fortunately this turned out to be an issue with the way with we were using picca, which was automatically projecting out long wavelength modes. Using a setting to prevent it from doing so solved the problem and the correlation functions now look much better!
Awesome, thanks for letting us know! I'll close this, but feel free to reopen if things still look weird.
@jfarr03 ran a correlation function measurement on the density skewers (using the public code picca), and found that the correlation function measured was not isotropic. Instead, it showed the usual trend of linear redshift space distortions (RSD), with the line of sight correlation crossing zero on smaller scales, and the transverse being positive up to very large separations.
This is present in both the skewers branch and in the revamp/master branch, and I managed to reproduce the results on a different simulation.
It is possible that this is caused by some post-processing of the skewers, or a wrong usage of the code to measure the correlation function, but before digging too much into that we wanted to hear your opinion, @damonge, @fjaviersanchez. Is it possible that the skewers are using some sort of redshift space density???