More realistic likelihood computations and marginalising over the nuisance parameters

[landmanbester]

Currently we compute the likelihood without marginalising the nuisance parameters. In most cases this will favour LCDM models since the astrophysical processes used to generate the data are typically calibrated under the assumption of an LCDM model. Once again realistic tests of the CP will require marginalising over the nuisance parameters.