CosmoMC: fnu & propose_matrix

[Xin Wang]

Hi, Antony!

It is found out that all the supplied propose matrices give zero correlation between fnu and other cosmological parameters. Given some solid degeneracy between the sum of neutrino masses and other parameters, like n_s, is it still ok to set one of the default covariance matrices, e.g. params_CMB.covmat, to be propose_matrix, when sampling chains where fnu is assigned a non-zero prior?

Thanks!

[Antony Lewis]

Yes, it will just take longer -make sure you set up the starting proposal width appropriately in the .ini file. Or use the suplpied planck_nu.covmat.

[Xin Wang]

Yes, it will just take longer -make sure you set up the starting proposal width appropriately in the .ini file. Or use the suplpied planck_nu.covmat.

Thanks, Antony! I found out that planck_nu.covmat only includes the (co)variances for seven parameters. If some other parameters are also assigned a non-trivial prior, do we have to insert certain lines & columns from other .covmat file, e.g. planck.covmat, into planck_nu.covmat to make a new propose matrix?

Furthermore, if the answer to above question is no, is it equivalent to say that the choice of propose_matrix does not affect the constraint results nor the parameter degeneracy, when the "start width" for each parameter is appropriate?