How to compute the "factorial root"?
David Wilson
davidwwilson at comcast.net
Wed Jun 4 02:19:08 CEST 2008
----- Original Message -----
From: "Alonso Del Arte" <alonso.delarte at gmail.com>
To: <seqfan at ext.jussieu.fr>
Sent: Tuesday, June 03, 2008 7:27 PM
Subject: How to compute the "factorial root"?
> So is there a formula [for the inverse of the gamma function], or will I
> just have to settle for a program that merely automates my trial and
> error?
>
> Alonso del Arte
No, there is no formula for inverse gamma. Your best bets are probably
bisection or the secant method, depending on how much work you want to put
into this. If you have a mathematical language on hand, you could probably
use a numerical root finder on Gamma(x)-y = 0.
But even if there were, it would not necessarily help you evaluate inverse
gamma any more easily or accurately. For example, there are formulae for the
roots of cubics and quartics, but they involve so many roots and divisions
that it's computationally faster and more accurate to use Newton's method to
compute them. And in some cases, a formula introduces numerical
instabilities that are avoided by "trial and error" methods.
Even numerical evaluation of functions as simple as 1/x and sqrt(x)
ultimately involve "trial and error" methods of varying sophistication. A
formula just hides the work involved.
More information about the SeqFan
mailing list