Bug #1795
ctulimit gets problems if parameter error is zero
| Status: | Closed | Start date: | 06/20/2016 | |
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assigned To: |  Knödlseder Jürgen | % Done: | 0% | |
| Category: | - | |||
| Target version: | 1.1.0 | |||
| Duration: |
Description
In some were cases, it may be that be parameter error is zero (either not fit was performed before, or something went wrong in the fit due to faulty inputs etc).
In any case, ctulimit computes the range for the limit calculation from the parameter error. If the error is zero, the range is not existent and the calculation doesn’t converge. I propose to either throw a meaningful exception if the range is zero, or to simply add some buffer to the range.
Recurrence
No recurrence.
History
#1
Updated by Knödlseder Jürgen over 8 years ago
I agree that this should be changed.
Have you an example of a fit where the parameter error is zero? This may point to another underlying problem.
#2
Updated by Mayer Michael over 8 years ago
Have you an example of a fit where the parameter error is zero?
Not at the moment. It may happen in case of a faulty user input (e.g. instrument for a model is set to CTA, but the data is only “HESS”). Then the fit will stay at its original values and the errors will be zero.
Another thing I remember (don’t have an example at hand though) was that this happened sometimes when parameters hit the limit (but this also was some time ago).
I guess simply adding a meaningful error message would be good to implement.
#3
Updated by Knödlseder Jürgen over 8 years ago
I propose to use simply the parameter value as parameter error in case that the error is zero. Remains to be seen whether this is actually sufficient.
#4
Updated by Knödlseder Jürgen over 8 years ago
- Status changed from New to Closed
- Assigned To set to Knödlseder Jürgen
Merged into devel.