Action #2456
make sure residual computation with SUBDIV method fails graciously for 0 counts
| Status: | Closed | Start date: | 04/20/2018 | |
|---|---|---|---|---|
| Priority: | High | Due date: | ||
| Assigned To: |  Tibaldo Luigi | % Done: | 100% | |
| Category: | - | |||
| Target version: | 1.6.0 | |||
| Duration: |
Description
Talvikki Hovatta reported failures of csrespec that seem to be related to bins with 0 counts. Make sure that in this case the tool fails graciously, i.e., still gives output with Nan in empty bins.
Recurrence
No recurrence.
History
#1
Updated by Knödlseder Jürgen almost 6 years ago
- Assigned To set to Tibaldo Luigi
- Priority changed from Normal to High
- Target version set to 1.6.0
#2
Updated by Tibaldo Luigi almost 6 years ago
- Status changed from New to In Progress
Looking for a test case to reproduce the issue ...
#3
Updated by Tibaldo Luigi almost 6 years ago
- % Done changed from 0 to 10
In fact it’s not csresspec that fails, but just the example plotting script. This is more logical since the SUBDIV method divides by the model (which is never exactly zero).
Here’s the error message, need to look into it
$CTOOLS/share/examples/python/show_residuals.py specres_SUBDIV.fits
Traceback (most recent call last):
File "/usr/local/gamma/share/examples/python/show_residuals.py”, line 341, in <module>
show_residuals()
File "/usr/local/gamma/share/examples/python/show_residuals.py”, line 329, in show_residuals
plot_residuals(args0, plotfile, hdu)
File "/usr/local/gamma/share/examples/python/show_residuals.py”, line 211, in plot_residuals
algorithm)
File "/usr/local/gamma/share/examples/python/show_residuals.py”, line 117, in fill_cmr
e_resid.append(err / c_model.real(row))
ZeroDivisionError: float division by zero
#4
Updated by Tibaldo Luigi almost 6 years ago
- Status changed from In Progress to Pull request
- % Done changed from 10 to 100
It does happen for WSTAT that the model is exactly 0 for the Off region.
The case is dealt with in obsutils.residuals, and therefore in csresspec, but it was not in the show_residuals example script when computing the residual error bars on the fly.
I modified it to set the error to 0 whenever the model is 0 for the SUBDIV and SUBDIVSQRT algorithms.
#5
Updated by Knödlseder Jürgen almost 6 years ago
- Status changed from Pull request to Closed
Merged into devel.