The fortran program "fitexy.f" is from "Numerical Recipes".
It computes linear regressions for data with errors in both
x and y.

The input data must be in /users/peter/nrfort/xy.dat, and
must have a particular format.  The first 5 entries are:

1) Npoints
2) 1
3) sigX
4) sigY
5) 0

They are followed by three columns of data, of with the
second and third columns are the x and y values.

Currently the program limits you to 10000 data points.
If you have more you have to change all occurrences of the
parameter NMAX.

The program "fitexy.f" uses fixed errors for x and y.  For
these errors I use 1X the measurement precision (in mixing
ratio units).  The program "fitexyz.f" uses different errors
for each point.  For that program the input data file
(/users/peter/nrfort/xyz.dat) is a little different.  The
first 3 entries are (1) Npoints, (2) 1, (3) 0, and there
are two additional data columns holding sigX and sigY.

The output goes to /users/peter/nrfort/results.NR.  The
format is:

  intercept    stderr
  slope        stderr
  CHIsquare    Q

(Q is the goodness-of-fit probability).