Skip to content

Arbitrary normalisation

Stefano Camarda requested to merge arbitrary-normalisation into master

This new feature is a generalisation of the "Normalised" option for datasets. The new code allows to normalise the prediction for a dataset to an arbitrary normalisation value. Previously the predictions could be normalised only to an unitary integral.

If used with a normalisation value equal to the integral of the data in all bins (multiplied by bin width), and in conjunction with an overall fully correlated uncertainty with 0 prior, it allows, in practice, to perform a fit to a normalised cross section with the need to modify the input data, nor the predictions.

In order to to perform a normalised fit the user needs to set the following in the data file:

NInfo = 1

CInfo = 'Normalised'

DataInfo = <totxs>

where <totxs> is the sum of the cross section in all bins of the dataset, multiplied by the bin width.

To have the usual behaviour for this option (predictions normalised to unitary integral), the previous settings are still valid:

NInfo = 1

CInfo = 'Normalised'

DataInfo = 1

Edited by Stefano Camarda

Merge request reports