Undefined behaviour of spectrum monitor if methods different from 'welch' or 'median mean' are set.

[mikelovskij]

It seems that in the init method (first chunk of code pasted) of spectrum.py and in the update data method where the averages are computed (third chunk of code pasted) , only the median mean and the welch method are implemented,
but in another part of the update_data method (second chunk of code pasted), other methods are also accepted (median, lal-...), but will be treated as if they were the welch method.

In init

weight = kwargs.pop('weight', 'linear')
if weight.startswith('exp') and self.method == 'median-mean':
    raise ValueError("Median-mean average PSD method is incompatible "
                     "with a non-linear weighting")

In update data (before the computation of the single ffts)

if (self.method in ['median-mean', 'median'] or
        self.method.startswith('lal-')):
    method = 'lal-welch'
else:
    method = 'welch'

In update data (at the end the averages are computed:)

if self.method == 'median-mean' and len(SPECTRA[channel]) == 1:
    spec = SPECTRA[channel][0].value
elif self.method == 'median-mean':
    odd = numpy.median(SPECTRA[channel][::2], axis=0)
    even = numpy.median(SPECTRA[channel][1::2], axis=0)
    spec = numpy.mean((odd, even), axis=0)
else:
    weights = self.weights[-len(SPECTRA[channel]):]
    weights /= weights.sum()
    spec = numpy.sum([s*w for (s, w) in
                      zip(SPECTRA[channel], weights)], axis=0)