It appears that the message board will not accept the spreadsheet attachment. Sorry.
Leigh _____ From: Leigh J. Halliwell [mailto:[EMAIL PROTECTED] Sent: Tuesday, May 08, 2007 6:05 PM To: 'Programming forum' Subject: RE: [Jprogramming] Digamma function Dear J Forum: Thank you all for your comments. Indeed, it is better to calculate lngamma from numerical recipies than to take the log of the J function. I compared J with numerical differentiation against R in the domain [0, 10]. (See attached spreadsheet.) Lngamma and digamma (first numerical derivative D.1) were good. Trigamma was fair, and tetragamma was nonsense. Trigamma was a little better if I defined it as D.1 of digamma than as D.2 of lngamma. In no way could a get a good calculation of tetragamma. This just goes to show that I'll have to program it from some numerical recipe to assure accuracy. Thanks. Sincerely, Leigh -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of R.A. MacDonald Sent: Tuesday, May 08, 2007 4:03 PM To: Programming forum Subject: Re: [Jprogramming] Digamma function Leigh; Do you have examples: x y pairs where you can test y = foo D.n x? I suspect that Roger et al have covered many of the usual suspects: polynomials, trig functions, even gamma... the J help page for D. seems to confirm this. Leigh J. Halliwell wrote: > Dear Raul and J Forum: > > I was aware of the D. operator, and admire the solution. My only question > is how accurately D. performs numerical differentiation. Can I rely on it, > or should I seek numerical methods for deriving the gamma functions? ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
