MFIT /STEP
[SIC\]MFIT Y=f(X,&a,...) /STEP A2 B2 ...
A2 B2 ... are used to pass the unity vectors (steps) for the iteration
on parameters &A &B (defaulted to 0.25). Poor choice of initial steps
may lead to non convergence. Too small steps will not help converging
when one starts too far from the solution. Too large steps may lead to
incorrect evaluation of the parameter errors. The optimal step for the
determination of the errors is about the error bar, so that all parame-
ters become dimensionless.