John Manslow wrote:
There are two tricks that are normally used to overcome the problem that
OpenOffice has with its implementation of the Poisson distribution. The
first is to perform all interim computations on the natural logs of the
quantities involved. This prevents large and small quantities from over
or underflowing. The second is to replace the factorial with the gamma
function or, since we're working with logs, the log gamma function.
Thanks a lot for your input.
To preserve accuracy, it seems best to keep using the old formula where
it doesn't overflow, and use yours otherwise.
I'd like to help further with this aspect of OpenOffice's functionality
but have very little spare time at the moment because I'm co-editing a
book in the evenings and at weekends. If there's any particular area
where you think I could help or if there are any other areas of
statistical functionality that developers are struggling with, please do
not hesitate to contact me and I'll see if I can help.
If you are aware of other functions that can be improved, help would be
very welcome. Some functions (like standard deviation) have been changed
for OOo 2.0, so it would be best to use a recent 2.0 snapshot build.
To make use of any code contributions, we need a JCA (Joint Copyright
Assignment, see http://contributing.openoffice.org/programming.html).
The changes mentioned above can be done without copying actual code, but
it would be great if you could submit the JCA for future contributions.
Niklas
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
