[Rd] Re: [R-SIG-Mac] Bug running pbinom() in R-GUI?

[Prof Brian Ripley]

This has no context attached, but I believe refers to a message of Feb 10 
(that I deleted locally a while ago).

R-devel uses a different version of pbeta, which has neither the problems 
mentioned for the Gnumeric version described in PR#7153 nor the slowness 
of the previous version.  From the NEWS file

     o	pbeta() now uses a different algorithm for large values of at
 	least one of the shape parameters, which is much faster and is
 	accurate and reliable for very large values.  (This affects
 	pbinom(), pf(), qbeta() and other functions using pbeta at C
 	level.)


On Tue, 22 Feb 2005, Morten Welinder wrote:

> The real problem is that pbeta can take forever.  That's bug #7153 and a fix is
> within reach.

I have never managed to make it take more than 18secs before it got an 
internal integer overflow on the loop size.  Multiply that by around 1000 
for the MacOS X version with its slow R_ProcessEvents, and 10 for the 
Windows version.

On the task in the original message, pbeta itself was taking a few 
milliseconds (hardly `the real problem'), all the rest of the time being 
taken by MacOS X R_ProcessEvents.


-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595