Opened 12 years ago
Closed 12 years ago
#161 closed defect (fixed)
W50 calculation can fail for negative searches
Reported by: | MatthewWhiting | Owned by: | MatthewWhiting |
---|---|---|---|
Priority: | highest | Milestone: | Release-1.2.1 |
Component: | Searching | Version: | 1.2 |
Severity: | critical | Keywords: | |
Cc: |
Description (last modified by )
When running negative searches, Attila found a bunch of sources with very small W50 values. I was able to reproduce this (see /DATA/CUSCUS_1/whi550/WALLABY/HIPASS2tests) - in fact some w50 values came out to be zero.
I think the problem is doing the parameterisation after the re-inversion of the cube. Perhaps the reinversion of the cube should be done after this, but still prior to any output of relevant arrays (such as saving the smoothed array).
Change History (4)
comment:1 Changed 12 years ago by
Description: | modified (diff) |
---|---|
Status: | new → assigned |
comment:2 Changed 12 years ago by
comment:4 Changed 12 years ago by
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
Closing - 1.2.1 released.
Note: See
TracTickets for help on using
tickets.
OK, found the problem, and a likely solution. The problem was that the data was being re-inverted, and so the velocity width calculations were using a negative spectrum. This buggered up the interpolation/extrapolation expressions, leading to spurious results.
A simple fix was to invert the integrated spectrum passed to the calculation function, so that the features it is working with are positive and the various assumptions hold. This is possible using the negSource member that is set when we do the full-scale reInvert() call.
Call this fixed, but need to test a bit more rigorously.