The 
statistic assumes that all the spectral channels are
Gaussian distributed. If there are small numbers of counts in a
channel this will not be true. An alternative statistic, the
C-statistic, can be used in this case provided that the data has
not been background subtracted. The C-statistic now in use does
provide a goodness-of-fit and this can be checked using the
goodness command which uses Monte Carlo methods. This C-statistic
can also provide confidence intervals in exactly the same way as .
So give the command stat cstat and then use the fit and
error commands as usual. An alternative approach is to continue
using the statistic but change the weighting to provide a
better approximation to a Gaussian distribution in the small number
limit. This can be done using the weight gehrels or weight
Churazov commands.