Feature #603
Perform a scientific validation of the P7V6 Fermi-LAT response
| Status: | Closed | Start date: | ||
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assigned To: | - | % Done: | 100% | |
| Category: | - | |||
| Target version: | 00-09-00 | |||
| Duration: |
Description
The P7V6 response has been implemented, but no scientific validation has been done with respect to the Fermi-LAT Science Tools.
GammaLib (or better ctools) should be applied to several test cases, and results should be compared with the ScienceTools. It would also be interesting to track executing speeds, and to compare the performances of GammaLib with the ScienceTools.
I’m also very interested to see how GammaLib deals with many sources in the ROI. This may help to improve the GammaLib optimizer.
Recurrence
No recurrence.
Related issues
History
#1
Updated by Knödlseder Jürgen about 12 years ago
- Description updated (diff)
#2
Updated by Knödlseder Jürgen almost 11 years ago
- Status changed from New to In Progress
The Wiki pages Validation of Pass 7 IRFs and Validation of Pass 7REP summarize the result of the comparison of ctlike runs and Fermi/LAT Science Tools gtlike runs. For the specific test case, a discrepancy exists between the fitted parameters. Specifically, the Crab flux is significantly different. Here the results for the analysis of one week of Fermi/LAT data:
| Parameter | gtlike | ctlike | gtlike (REP) | ctlike (REP) |
| Crab counts (for same flux) | 814.987 | 842.551 | - | - |
| Flux [1e-7] | 24.23 +/- 1.50 | 19.23 +/- 1.21 | 25.59 +/- 1.51 | 19.92 +/- 1.19 |
| Spectral index | 2.125 +/- 0.048 | 2.039 +/- 0.051 | 2.135 +/- 0.046 | 2.051 +/- 0.049 |
| Extragalactic diffuse | 2.286 +/- 0.462 | 2.896 +/- 0.442 | 2.876 +/- 0.686 | 3.486 +/- 0.694 |
| Galactic diffuse | 0.862 +/- 0.065 | 0.863 +/- 0.064 | - | - |
| Galactic prefactor | - | - | 0.729 +/- 0.144 | 0.774+/- 0.157 |
| Galactic index | - | - | 2.876 +/- 0.686 | 3.486 +/- 0.694 |
The first line “Crab counts (for same flux)” is interesting as it gives the number of counts predicted for the Crab nebula for a given flux. This indicates that ctools/GammaLib produces more counts for a given flux compared to gtlike. Correspondingly, for a given number of counts, the ctools/GammaLib is lower.
It remains to be seen where this difference comes from.
Here a list of tests that are possible:- determine the number of Crab counts for a given flux for gtlike and ctlike for narrow energy bands and as function of energy
- determine the number of Crab counts for a given flux for gtlike and ctlike for an increasing number of energy bins
- fix the diffuse components
- simulate a source without background using gtobssim and fit results using gtlike and ctlike; could be done for various spectral indices and energy binnings
- check if the spatial binning has an impact
#3
Updated by Knödlseder Jürgen almost 11 years ago
- Status changed from In Progress to Closed
- Target version set to 00-08-02
P7V6 is now becoming outdated, so we may close this issue and concentrate on finishing #1004 which deals with P7REP.
#4
Updated by Knödlseder Jürgen over 10 years ago
- Target version changed from 00-08-02 to 00-09-00
- % Done changed from 0 to 100
Skip 0.8.2 release.