Problems with the covariance matrices for large values of kmax

[andreufont]

When I run analyses on mock data with kp < 0.5 1/A I don't have any problem, but if I run with kp < 1 1/A I get warnings with:
Det(cov) appears to be 0: could be a singular covariance matrix! iz = 1, it_A = 5

This shouldn't happen... we should investigate what's going on. I would have imagined that at high-k things should be well behaved, but it might be that the rebin factor of 3 is not enough, and that we should use a rebin factor of 4.

For instance, I'm seeing this problem with iz=1 for these files:

tru_cont/tru_cont_binned_out_bf3_px-zbins_4-thetabins_10_w_res_avg50.hdf5

uncontaminated/uncontaminated_binned_out_bf3_px-zbins_4-thetabins_10_w_res_avg50.hdf5

contaminated/contaminated_binned_out_bf3_px-zbins_4-thetabins_10_w_res_avg50.hdf5

[andreufont]

@sindhusatyavolu - could you test this? You don't need to run fits, just check that the rebinned covariances are well behaved, for different mocks and different kmax.

For instance, you could do this:

for iz in range(N_z):
    for it_A in range(N_A):
        det_cov = np.linalg.det(data.cov_ZAM[iz, it_A,:,:])
        if det_cov <= 0:
            print("Det(cov) non-positive for iz = {}, it_A = {}".format(iz, it_A))