Error in BAO/RSD Likelihood with CLASS

[jcolinhill]

Running the default SDSS DR12 BAO+RSD likelihood when using CLASS as the theory code produces the following error:

[9 : bao.sdss_dr12_consensus_final] Initialising.
[9 : classy] ERROR Requested product not known: {'fsigma8': {'z': array([0.38, 0.51, 0.61])}}

It looks like the interface to the CLASS computation of f*sigma8 is broken, which is probably an easy thing to fix.

Thanks for this great code!

[JesusTorrado]

Hi @jcolinhill

fsigma8 is so far not implemented in CLASS. I did let their developers know about it a while ago, and I have been told it may come in the next version, 2.8, in a few days (with the caveat that they disagree about it's usefulness, given that it assumes a scale-independent growth factor).

[cmbant]

CAMB defines f sigma_8 so that it doesn't make that assumption following Eq 33 of Planck params 2015. (whether that's actually what is measured by surveys is another question..)

[jcolinhill]

Hi @JesusTorrado @cmbant - thanks for the quick reply. Do you know if the updated CLASS version will be out very soon? If not, perhaps I'll try implementing the P15 prescription within CLASS...

[JesusTorrado]

I have been told that it will be released in the next few weeks as CLASS 2.8.

If you want to make sure that fsigma8 is added in the new version, the best thing you can do is contacting CLASS' developers, and maybe offer them a pull request (but I would contact them first anyway, since they told me that they would probably include it themselves).

[lukashergt]

Hi @jcolinhill and @JesusTorrado,

any update on this?

I believe classy now has a function called scale_independent_growth_factor_f in class_public/python/classy.pyx.
Could that be used in Cobaya's CLASS version of get_fsigma8 in cobaya/theories/classy/classy.py?

[JesusTorrado]

Hi @lukashergt

I expect to be taking a look at this in the next few days. Once the German semester is over (~ Feb 15th) we can try to get Julien Lesgourgues involved too.

[JesusTorrado]

scale_independent_growth_factor_f appears to be one piece of the thing, but we would still need the density-velocity correlation (see https://arxiv.org/pdf/1502.01589.pdf). Hopefully @lesgourg can add it? There is an issue open already at CLASS: lesgourg/class_public#393