Feature #2662

ctlike/ctulimit tolerance

Added by Sokolenko Anastasia about 6 years ago. Updated about 6 years ago.

Status:ClosedStart date:08/15/2018
Priority:NormalDue date:
Assigned To:Knödlseder Jürgen% Done:

100%

Category:-
Target version:1.6.0
Duration:

Description

We noticed that for a weak dark matter signal above CR background, ctlike and ctulimit return results too good to be true in comparison with a simple signal/noise estimation for 500 hours of simulated data.
Artificial reduction of time (to have no more than 1e4 photons in 10% broad energy bin) seems to solve the issue. Therefore, the issue connected to numerical accuracy.

It seems that in some cases, especially for the long observational time, default fitting/integration precision is not sufficient.
It will be good to be able to control the fitting/integration tolerance with e.g. the additional hidden parameter for ctlike/ctulimit even by the price of a potentially significant increase of the computational time.


Recurrence

No recurrence.

History

#1 Updated by Knödlseder Jürgen about 6 years ago

  • Assigned To set to Knödlseder Jürgen
  • Target version set to 1.6.0

We can add the fit tolerance as a hidden parameter. It would be good if you could provide a specific test case that reproduces your problem to see whether fit tolerance is actually the issue.

#2 Updated by Knödlseder Jürgen about 6 years ago

  • Status changed from New to Pull request
  • % Done changed from 0 to 100

I added a like_accuracy parameter to all tools and scripts that do maximum likelihood fitting. I will now merge the code into devel.

#3 Updated by Knödlseder Jürgen about 6 years ago

  • Status changed from Pull request to Closed

Code is integrated in devel branch.

Also available in: Atom PDF