steppar

       plot 1D or 2D chi^2 contours for chi^2 fitting error determination

       fitting

       steppar,x,y,w,a plus lot's of keywords

       x,y,w: (x,y) values and weight of data, identical to
              curvefit
       a: best fit parameters of the fit function

     * required:
         par1ind: index of 1st parameter to be stepped
         par1min,par1max: min. and max. value of 1st parameter
         nstep1: number of steps for par1
     * optional:
         par1log: if set, use logarithmic steps
         par2ind,par2min,par2max,nstep2,par2log: same as
            the ..1.. pars for the 2nd parameter to be stepped.
         chi2red: reduced chi^2 at minimum (result from chi2 keyword
                of curvefit)
         dof: degrees of freedom (returned from curvefit)
         plot: if set, a plot of chi^2 vs. the parameter(s) is
            shown, with contours indicating the 1sigma, 90%, and 99%
            confidence regions (the latter option requires the
            chi2red and dof values as well!).
         sigmaa: if also given (returnvalue from jwcurvefit) also
            plot the error bars as determined from the Hessian
            matrix (also need chi2red and dof).
         dev: plot axes as deviation from best fit value
       * All keyword-parameters given to the previous call to
         curvefit.

       par1val: values of 1st parameter
       par2val: values of 2nd parameter
       chi2val: array containing the chi^2 values

       none

       if /plot is set, a (contour) plot is drawn in the current window

       stepping regions need to contain the best fit value.

   (from a program doing a fit to an astrophysical absorption line)
   bounds=[ $
            [0.,10000.], $
            [xstart,xend], $
            [0.,20.], $
            [0.,10.], $
            [0.,0.] $
          ]
   ;; do the fit
   res=jwcurvefit(lam,dat,w,a,sigmaa,function_name='hbetaline',/noderivative, $
                  chi2=chi2,itmax=200,tol=1D-5,iter=iter,bounds=bounds, $
                 dof=doffit)
   steppar,lam,dat,w,a,$
     par1ind=1,par1min=a[1]-2.,par1max=a[1]+2.,nstep1=30, $
     par2ind=2,par2min=a[2]-0.75,par2max=a[2]+0.75,nstep2=30, $
     par1val=par1val,par2val=par2val,chi2val=chi2val,$
     chi2red=chifit,dof=doffit,$
     FUNCTION_name='hbetaline',/noderivative,bounds=bounds,/plot

     Version 1.0, 1999/03/03, Joern Wilms ([email protected])
     Version 1.1, 1999/11, 23, JW:
         added psym plotting option,
         removed bug in computation of fit uncertainty (dof was
            overwritten)
         added option to also plot uncertainty from Hessian matrix

[Home Page] [Software, Documentation] [IDL Documentation] [Quick Reference] [Feedback]