To use likelihood fitting with ROSAT data, you proceed in just the
same way as for chi-squared fitting:
XSORT to get source and background spectra
XRTSUB to subtract background, use BGMODEL=bfile to
generate a background file for spectral fitting
XRTCORR corrects the spectrum to ct/s (no need to
correct the bfile)
XRTRESP attaches response to spectrum
SFIT LIK will perform likelihood fit, picking up bfile
automatically
SERROR LIK generates confidence region using Cash statistic
SPLOT as normal
Likelihood fitting will also work with spectral sets. The background
file must in this case be a spectral set of the same dimensions as
the source dataset (as it will be if you follow the above procedure).