Re: tolerance in binopdf.m

[Ben Abbott]

On Sep 21, 2011, at 2:56 AM, Michael D Godfrey wrote:

> I just checked ML.  They reference Abramowitz and Stegun
> and C. Loader.  And, the details look OK.  
> 
> I do not have access to an ML version on Mac to see
> if it fails.
> 
> Also, I have not figured out the values of n for which
> the Mac fails, but in any case the Loader algorithm should
> be used for large n. (n=>10)
> 
> Michael
> 

For gammaln, Matlab on MacOS gives the same answer as Octave.

fprintf ('%.17f\n', gammaln (3))
0.69314718055994540
fprintf ('%.17f\n', log (0.5))
-0.69314718055994529

I don't have Matlab's binopdf() function to compare to.

Ben