KeyError when writing kappa to outfile after thresholding with concordant=False

When thresholding with concordant=False (not passing -C/--concordant) and also passing -k/--kappa_confusion_outfile or calling kappa() on the thresholded Dataset d, we get the following error:

...
  File "...\lib5c\tools\threshold.py", line 133, in threshold_tool
    kappa_dict = kappa(d)
  File "...\lib5c\algorithms\thresholding.py", line 182, in kappa
    for cond in d.conditions}
...
KeyError: ('rep_significant', 'ES_Rep1')

This happens because under concordant=False, final significance calls are never made on a per-rep level, and only exist on the per-condition level.

One alternative is to fall back to 'rep_significant_unfiltered' which is always guaranteed to exist. Another alternative is to say this combination is not supported, but this is a little strange because it does occur in the default pipeline.

Comments (1)