Pull distribution of diffuse cube model is biased
|Assigned To:||Knödlseder Jürgen||% Done:|
I recognized that I had not yet tested the diffuse cube model, and indeed the pull distribution is biased.
The corresponding pull evolution plots (mean and standard deviation of pull as function of number of samples; the mean should tend to zero while the standard deviation should tend to one) are attached as the first four plots (unfortunately I used the same names later for the pull distributions and I cannot undo that; hence click on the plots to see).
#2 Updated by Knödlseder Jürgen over 6 years ago
- File cube_bkg_index.png added
- File cube_bkg_prefactor.png added
- File cube_index.png added
- File cube_prefactor.png added
- File fine_bkg_index.png added
- File fine_bkg_prefactor.png added
- File fine_index.png added
- File fine_prefactor.png added
- File large_bkg_index.png added
- File large_bkg_prefactor.png added
- Status changed from New to Pull request
- % Done changed from 0 to 100
I did a couple of pull distributions to see whether things are okay now. Below the results. The top row was for a Gaussian map with sigma decreasing as function of energy (the Gaussian is much smaller than the ROI). The mid row shows the same but for a Gaussian map that is larger than the ROI. And the bottom row shows for a model with 100 energy layers instead of the 10 used for the models before. Everything looks fine.