is beta_qso shifted

[londumas]

This is another ticket, that could also be posted to picca, because we don't really know which is producing that.
The expected value of beta_qso, according to the file Bz_qso_G18.txt (that migh have changed @jfarr03?), and to the LCDM cosmo:
beta_QSO(z=2.310) = 0.25798.
We measure a beta_QSO a bit different.
combined: <beta_QSO> = 0.24695 +/- 0.00092 (<0.0027>, 0.00086).
It is totally possible that this systematic error is linked to picca, it has never been tested.
The best way to make sure is to compute the auto-correlation of quasars.
Anyway it is close enough to have the same effect.

mock = {}
mock['input_bias_QSO_evol'] = '$HOME/Run_programs/igmhub/picca/CoLoRe_mocks/v4.0/Bz_qso_G18.txt'

def alpha(z,a0,a1,zeff=2.30):
    evol = a0*((1.+z)/(1.+zeff))**a1
    return evol
def growthRateStructure(z, omega_M_0=0.3147):
    omega_m = omega_M_0*(1.+z)**3 / ( omega_M_0*(1.+z)**3+(1.-omega_M_0))
    f = sp.power(omega_m,0.55)
    return f

biasCoLoRe = sp.loadtxt(os.path.expandvars(mock['input_bias_QSO_evol']))
betaCoLoRe = growthRateStructure(biasCoLoRe[:,0], omega_M_0)/biasCoLoRe[:,1]
fromztobiasQso = interp1d(biasCoLoRe[:,0],betaCoLoRe)

[londumas]

It can also simply be related to the difference in the correlation function at small scales, as shown in this ticket #3.
On a similar note, sending the combined fit to the stacks, with a free growth_factor, I get:

beta_QSO = 0.379 +/- 0.0135
growth_rate = 1.39 +/- 0.0399

[londumas]

Using the Lya absorption in the Lyb region, we get the same results:

[londumas]

The following plot was requested by @andreufont and shows the evolution of the best fit beta_qso in the combined fit of the stack of 10 auto + 10 cross (Lya+Cont+Noise), as a function of the minimum distance of the fit: rmin. The different lines show the different values expected for effective redshift (orange), the minimum and maximum redshift of the bins of the auto (green and red) and the same for the cross (violet and brown).
The fact that the first two points are outside the zmin,zmax range really shows that the small scale clustering is not the same as the theory, however the plateau is between min and max and thus shows that the apparent systematic error can come from the definition of zeff in picca, and thus is less troubling.
We don't have to fix the small scales, but it is nice to know that it is like that.
The QSO auto-correlation would give better idea.

[andreufont]

That's great!

If you worried about z_eff, you could measure the cross-correlation on a small redshift range (say 2.3 < z < 2.5), but I would do this type if testa on the transmission files where the measurement is less noisy.