Adapt SensitivityEstimator to support MapDataset and joint datasets about gammapy HOT 4 OPEN

Hi @registerrier ,

as I mentioned to you this approach is working quite well for SWGO after I adapted it to our code base.

We will now add the experimental implementation to our data processing framework (unfortunately still not public, but hopefully soon).

Do you think we should discuss how to incorporate it here?

I guess the minimum would be:

• the function to compute the cash statistics (which is the one used by the map datasets) from the asimov dataset
• the class for the TS defined by Cowan
• a new MapSensitivityEstimator which I guess can be the one we are using but further improved (what is currently missing e.g. is the possibility to define a minimum number of excess gamma rays and add the significance criteria to show where this is dominated by excess or background systematics; the former is more negligible for instruments like SWGO than CTA, but I think it should be added also)

What do you think?

from gammapy.

I guess the minimum would be:

• the function to compute the cash statistics (which is the one used by the map datasets) from the asimov dataset
• the class for the TS defined by Cowan
• a new MapSensitivityEstimator which I guess can be the one we are using but further improved (what is currently missing e.g. is the possibility to define a minimum number of excess gamma rays and add the significance criteria to show where this is dominated by excess or background systematics; the former is more negligible for instruments like SWGO than CTA, but I think it should be added also)

@HealthyPear That sounds like a good plan, if you already have a working solution you are welcome to open PRs for each of these points.

from gammapy.

Sorry coming back to this after a while....

I can open a draft PR with the 3 ingredients described above:

• I plan to add the MapSensitivityEstimator under gammapy.estimators.maps
• the rest under a new module gammapy.stats.asimov

would that be OK as a start or you had in mind a different structure?

from gammapy.

As reminder, see also the project https://github.com/gammapy/gammapy-sensitivity

Do we have a reasonable global design?

from gammapy.