High Energy Bias at large offsets

[matthew-w-lundy]

When using the full enclosure IRFs and doing a likelihood analysis, the spectra at higher energies all show a negative bias (i.e. lower flux values). The bias can be as great as ~40%. Does not appear to significantly affect 0.5 degree offset.

[steob92]

Ok so brief hiccup before the morning coffee kicked in...

Can you provide a minimal reproducible example, or link me to a notebook on the cluster or provide the runlist you're using? I cannot reproduce this error with:

• ED v491
• Cuts: Hard2Tel
• gammapy 1.3
• Python 3.12.3

I'm using 20 crab runs taken at 0.7 degrees between 2018-2020. I don't see any significant bias at high energies. There are some discrepancies at lower energies,this is in the regime where we'd expect theta-cuts to have an impact. I'd suspect it would be within the systematic uncertainty of the PSF. Nonetheless I don't see a negative bias for full enclosure at high energies:

The "best fit model" referred to is just the magic_lp model used as a constant reference point.