MFIT

         [SIC\]MFIT Yvar=Func(Xvar,&A,&B,...) [/START A1,B1,...]
     [/STEP A2,B2,...] [/EPSILON e] [/WEIGHTS w] [/METHOD m]
             or
         [SIC\]MFIT    Filename     [/START A1,B1,...]     [/STEP A2,B2,...]
    [/EPSILON e] [/WEIGHTS w] [/METHOD m]


    Performs a least squares fit  using  the  SIMPLEX  (default)  or  POWELL
    algorithm  (if option /METHOD P is selected) from Numerical Recipes, see
    Press et al. for details. The least square fit tries to adjust  function
    Func of variable Xvar and parameters &A, &B, ... to match variable Yvar.

    Any function or combination of functions  known  by  SIC  may  be  used.
    Blanks are supported in the formula. The '=' sign is required.
    /START A1,B1,... (comma-separated list WITHOUT BLANKS) is used  to  pass
    starting guesses for the parameters &A &B (defaulted to 0.1),
    /STEP A2,B2,...(comma-separated list WITHOUT BLANKS) is used to pass the
    unity  vectors  (steps) for the iteration on parameters &A &B (defaulted
    to 1.0).

      Fit parameter DUMMY names are &A...&Z . Already defined SIC  variables
    may  be  used  in  the formula as long as they don't match the variables
    used by MFIT. The resulting  parameters  will  be  stored  in  the  real
    variable  MFIT.   Option  /EPSILON  can  be  used to specify the desired
    precision, which means the deviation of the mean squared  difference  of
    two  fit  iterations.   This  might NOT be the absolute error of the fit
    parameters.  An array with weights may be given by  option  /WEIGHT.  It
    must  match  the  dimension  of the formula result (defaulted to uniform
    weight).

    The formula can also be provided  by  a  disk  file  (default  extension
    .GRF),  where  lines  beginning  with an exclamation mark are treated as
    comments.  Inside the file, blanks can be used in the formula, which may
    be spread over several lines.

     The command MFIT uses the following global SIC variables:
           MFIT     : Fit parameters
           MFITMATH : Text string containing the formula with replaced dummies
           MFITOUT  : Found approximation (the application of MFITMATH)
           MFIT_RES : Residuals  (i.e. MFITOUT-Yvar)
           MFIT_SUM : Mean squared residuals
           MFIT_ERR : .FALSE. on successful completion.
    Suppose you read X,Y=f(X) and Z (errorbar on Y) in the 3 arrays X, Y and
    Z.  You could change Z into weights (for example SIC\LET Z 1|Z^2 ), then
    fit a cubic polynomial regression in variables X,Y  by typing
      MFIT Y=(&A*X^3+&B*X^2+&C*X+&D) /WEIGHTS Z
    The MFIT command will print the used formula:
    I-MFIT,  Formula (stored in variable MFITMATH) :
    (MFIT[01]*X^3+MFIT[02]*X^2+MFIT[03]*X+MFIT[04])
    The array MFIT will contain MFIT[1]=value of '&A', and so  on...  To  be
    used afterwards.