Only a single input dataset can be entered - if multiple datasets are to be fitted a file containing references to them must be first set up using SDATA.
A fixed number of iterations are performed (default 30) but if a minimum is not found the user may restart fitting by typing SFIT again (no arguments should be required). The current model parameters are stored in the fit_model data object, this is updated by the program every NUP (default 3) iterations. If the user is running under DCL the program can be aborted with CTRL-C at any point and the last update to the parameter values will be retained, these can then be used to generate plotted output or taken as the starting point for further fitting. Aborting under ADAM can currently only be effected with CTRL-Y, which also exits the monolith - this situation should improve soon.
Approximate error estimates for the free parameters are computed from the matrix of second derivatives at the minimum. These can be taken as only a very rough (factor of two or so) guide to the errors. Refitting at offset positions in parameter space is required for an accurate error assessment - this is performed by SERROR.
The error assessment is rather lengthy and can be disabled by invoking:
however, this is unwise if SERROR is to be run subsequently, since the latter program uses the approximate error values as a starting point.ICL> sfit noerr
The fit results are normally printed at program termination, but may be appended to file FIT.OP if the APPEND option is selected. This is useful when performing a sequence of related spectral fits. SERROR appends to the same file by default (which is created if necessary), although the filename can be changed in both programs.
A source redshift may be included in the fitting process by specifying Z=value on the command line. The specified redshift will be used thereafter (also by SERROR, SPLOT and SFLUX) until disabled. The fitted model in this case corresponds to the spectrum in the SOURCE frame, and is redshifted by the software before being folded through the instrument response.
Likelihood fitting is available by including LIK on the command line.