Support #2655
negative slope for CR
| Status: | Closed | Start date: | 08/01/2018 | |
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assigned To: | - | % Done: | 0% | |
| Category: | - | |||
| Target version: | - | |||
| Duration: |
Description
We were doing a simulation of a cosmic ray background with a spectral model given by a constant (i.e. powerlaw with a slope=0) and fitting the simulated data with ctlike assuming the slope to be free. We found, that the best-fit slopes are systematically negative for 200 random realizations of CR data.
May you take a look on this issue, please? Potentially it can lead to biased estimation of the slopes of any other sources present in the model.
We checked, that this issue is present in ctools 1.5.2.
indexes.pdf
(11.5 KB)
biasedCRslope.tgz (11 KB)
Recurrence
No recurrence.
History
#1
Updated by Knödlseder Jürgen over 5 years ago
- Tracker changed from Bug to Support
My suspicion is that this is simply the numerical accuracy of the code. You are comparing pretty tiny numbers here. The difference between observed and predicted counts is 0.003, hence a relative precision of 6.3e-10!
But the difference is not zero, hence there is of course an error/bias somewhere. I guess you will also see a bias in the Prefactor.
You can always increase numerical accuracy at the expense of speed. In other words, there is a trade-off between accuracy and speed, and the trade-off was placed so that astrophysical results should never be worse than 1% (they often are much better).
#2
Updated by Knödlseder Jürgen almost 4 years ago
- Status changed from New to Closed