Background2D running on IRF with low statistics (few bins only) gives very different results between 1.0 and 0.17 #4345
Unanswered
vkulikovskiy
asked this question in
Help
Replies: 2 comments
-
The default |
Beta Was this translation helpful? Give feedback.
0 replies
-
Thanks for support. In fact the value of integral becomes:
which is much closer |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Behaviour:
v0.17 - nearest values are not 0 in "evaluate" with "nearest"
v1.0 - nearest values are 0 in "evaluate" with "nearest" and few orders of magnitude lower than in 0.17 for "evaluate+nearest" and "integrate_log_log" vs "evaluate_integrate".
It looks something fishy may happen for the evaluation on the 2D grid, i.e. for 0.17 non 0 value is searched while 1.0 is just nearest. It could be that grids for some reason are not the same (although axis check does not reveal it).
It is not clear if such behaviour is known and it is correct in v1.0 (but then wrong in 0.17?).
Minimal example:
0.17
Output:
v1.0
Output:
Here is the fits file (the extension is .fits, I've added .txt just to upload it here).
bkg_mu.fits.txt
Beta Was this translation helpful? Give feedback.
All reactions