.TH CURFIT 3M
.SH NAME
curfit
.SH DESCRIPTION
subroutine curfit.f 

source
  Bevington, pages 237-239.

purpose
  make least-squares fit to a non-linear function
    with a linearization of the fitting function

usage 
  call curfit (x, y, sigmay, npts, nterms, mode, a, deltaa,
    sigmaa, flamda, yfit, chisqr)

description of parameters
  x      - array of data points for independent variable
  y      - array of data points for dependent variable
  sigmay - array of standard deviations for y data points
  npts   - number of pairs of data points
  nterms - number of parameters
  mode   - determines method of weighting least-squares fit
           +1 (instrumental) weight(i) = 1./sigmay(i)**2
            0 (no weighting) weight(i) = 1.
           -1 (statistical)  weight(i) = 1./y(i)
  a      - array of parameters
  deltaa - array of increments for parameters a
  sigmaa - array of standard deviations for parameters a
  flamda - proportion of gradient search included
  yfit   - array of calculated values of y
  chisqr - reduced chi square for fit 

subroutines and function subprograms required 
  functn (x, i, a)
    evaluates the fitting function for the ith term
  fchisq (y, sigmay, npts, nfree, mode, yfit) 
    evaluates reduced chi squared for fit to data
  fderiv (x, i, a, deltaa, nterms, deriv)
    evaluates the derivatives of the fitting function
      for the ith term with respect to each parameter
  matinv (array, nterms, det) 
    inverts symmetric two-dimension matrix of degree nterms
      and calculates its determinant

comments
  valid for nterms up to 10
  set flamda = 0.001 at beginning of search