No ignition is causing infinite error

[rwest]

We have a simulation that does not trigger an ignition (the temperature never rises quite enough) and end up with 'average error function' of Inf and 'average deviation function' of -Inf. for the whole data set.

Looking at

PyTeCK/pyteck/eval_model.py

Line 405 in bba7984

error_func = numpy.nanmean(error_func)

        # calculate error function for this dataset
        error_func = numpy.power(
            (numpy.log(ignition_delays_sim) -
             numpy.log(ignition_delays_exp)) / standard_dev, 2
             )
        error_func = numpy.nanmean(error_func)

I'm guessing the nanmean is supposed to take a mean excluding NaN values, but the result of numpy.log(ignition_delays_sim) is not NaN but is -Inf when no ignition is detected, because process_results() sets ignition delay to zero if no ignition:

  else:
            warnings.warn('No ignition for case ' + self.meta['id'] +
                          ', setting value to 0.0 and continuing',
                          RuntimeWarning
                          )
            self.meta['simulated-ignition-delay'] = 0.0 * units.second

So we end up with 'average error function' of Inf and 'average deviation function' of -Inf.

Should we exclude infinities before averaging? include them somehow?
I guess technically it is an infinite error, but that doesn't help you rank things!

[rwest]

Just noticed this is a duplicate of issue #1 :)