search

sofia-calgaro / ZeroNuFit.jl

Bayesian unbinned fit of the neutrinoless double-beta decay
https://sofia-calgaro.github.io/ZeroNuFit.jl/
MIT License
2 stars 1 forks source link

new peak shape for MJD #51

Open sofia-calgaro opened 6 days ago

sofia-calgaro commented 6 days ago

Implement gaussian+low-E side tail for MJD partitions. Idea: add a key in the partition.json (under fit_group?) for specifying the peak shape to use for partitions listed in that file, eg "signal_shape": "gaussian" or "signal_shape": "gaussian_plus_lowEtail". In this way, we can handle differently different experiments

@tdixon97 any opinion?

tdixon97 commented 6 days ago

Yes of course that works, although it becomes a bit 'hardcoded' but idk a better way to do this..

sofia-calgaro commented 6 days ago

the only not-hardcoded way would be via the config.json, but I think there you lose some freedom... There you could specify something like:

        "events": [
            "legend-0vbb-config/events_gerda.json",
            "legend-0vbb-config/events_l200.json",
            "legend-0vbb-config/events_mjd.json",
        ],
        "partitions": [
            "legend-0vbb-config/partitions_gerda.json",
            "legend-0vbb-config/partitions_l200.json",
            "legend-0vbb-config/partitions_mjd.json"
        ],
        "signal_shape": [
            "gaussian",
            "gaussian",
            "gaussian_plus_lowEtail"
        ],

Anyway, we can set the gaussian as the default shape and let the user specify only when the other shape has to be used

tdixon97 commented 6 days ago

yeah i meant that ideally one could input directly a formula, but its very hard to implement, i think it belongs in the partitions file not the config file?

tdixon97 commented 6 days ago

so i agree with your original suggestion, its just a pity that you have to rely on names for shapes, but idk a way around it (similar for the background shapes).

sofia-calgaro commented 5 days ago

the new peak shape works after https://github.com/sofia-calgaro/ZeroNuFit.jl/commit/3d46a628329bc30a52990b67930e760e57425866 and was tested

However, in a combined fit, the fit does not work if we change the order, eg if we fit L200+MJD (this is related to the number of partitions for which we want to include the prior gamma ... this has to be handled better...)