search

VERITAS-Observatory / gammapy-tools

Repository with tools for gammapy analysis
GNU General Public License v3.0
2 stars 0 forks source link

Broad significance distributions #43

Open samanthalwong opened 4 months ago

samanthalwong commented 4 months ago

There was previously a bug where the we were correcting for the pileup at zero in significance distributions by plotting the OFF distribution as significance_off[exclusion_mask.data.flatten()]. When a gaussian distribution is fit to this distribution, it is broad (sigma ~ 1.15) and skewed negative (mu ~ -0.1) for several bright point source fields that show nice distributions in ED (Crab, Mrks).

It looks like some of the ON region is leaking into the background a bit.

This might be related to the energy range supplied to the ExcessMapEstimator (I get marginally better results by restricting the range of energies that are accepted), but I'm not sure.

I get identical results with and without bkg smoothing as well, so unrelated to that.

samanthalwong commented 4 months ago

e.g., for the 4 crab test runs: image

samanthalwong commented 2 months ago

This seems to be entirely related to the bkg generation - running the same pipeline with H.E.S.S. data gives perfect distributions.

So probably unrelated to ON region size, energy ranges, etc.