First, please note that the published version of the paper has many
typographical errors in the equations. This was due to unfortunate
mistakes in the type-setting stage that were not caught.
However, the preprint is fine. It can be found at:
http://www.astro.wisc.edu/~mab/archive/stats/stats.html. 

Second, the BCES code claims to handle covariance in X1 and X2.  The
equations are correct, but I did not code this correctly.  The BCES
code is fine if there is no covariance. But if you have covariance, it
really should be fixed. This is not hard to do but will take a little
bit for me to just to try and remember what's what!

