On Fri, 19 Jan 2007, Malcolm J. Currie wrote:
> Oh dear. As you gathered from my previous message, a missing root 2
> or inverting the 1.17 factor were the most likely reasons. I'm not at
> all surprised. If you want to use this stuff, a code walkthrough
> would be prudent, but might damage your health.
At least that would explain why people don't trust it at all. Still
doesn't meet the requirements of being able to constrain parameters and
fit negative gaussians so we still need a FITBEAM (and it seems that the
gaussian fitting in cupid is using a robust algorithm).
I'll patch that calculation in ESP at least.
>
>> so, um, only broken since 1997 then....
>
> It might have been in the original code, and Norman hadn't noticed the
> error (the words wood and trees spring to mind). ESP had become a
> poisoned chalice. My first feedback to Grant went on for page after
> page after page...
>
I'm pretty sure that bug has been in since 1997 (Norman took over in
1998).
--
Tim Jenness
JAC software
http://www.jach.hawaii.edu/~timj
|