search

astropy / specutils

An Astropy coordinated package for astronomical spectroscopy. Maintainers: @rosteen @keflavich @eteq
http://specutils.readthedocs.io/en/latest/
166 stars 127 forks source link

Possible bug: correlation.template_correlate throws an error if given Spectrum1D objects with no flux uncertainties defined. #1117

Closed scfleming closed 8 months ago

scfleming commented 8 months ago

OS: Mac 12.7.1 Python: 3.11.6 specutils: 1.11.0

Issue: I am attempting to use the template_correlate() function like so: corr1, lag1 = correlation.template_correlate(obs_spec_loglin, model_1_loglin, resample=False)

Here, obs_spec_login and model_1_loglin are both Spectrum1D objects. They do not have flux uncertainties defined. I run into the following error, which appears to be due to line 245, part of _normalize(), not handling the case where flux uncertainties are not assigned to the Spectrum1D object?

File "/Users/fleming/anaconda3/envs/py311/lib/python3.11/site-packages/specutils/analysis/correlation.py", line 77, in template_correlate normalization = _normalize(observed_log_spectrum, template_log_spectrum) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/fleming/anaconda3/envs/py311/lib/python3.11/site-packages/specutils/analysis/correlation.py", line 245, in _normalize unc = observed_spectrum.uncertainty.represent_as(StdDevUncertainty) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ AttributeError: 'NoneType' object has no attribute 'represent_as'

scfleming commented 8 months ago

I'll note the example use of this function in RTD does define a flux uncertainty in the Spectrum1D objects it creates as a demonstration, but one should be able to calculate cross-correlation without flux uncertainties defined.

rosteen commented 8 months ago

I think you're right that this is a bug - it looks like we weight the numerator and denominator in the normalization calculation with the uncertainty, but in the case that there is no uncertainty we just shouldn't weight it, not fail entirely.