F distribution on the 41C

[Dieter]

(12-22-2015 09:34 PM)Paul Dale Wrote:  [ Implementing the regularized Beta function ]
I don't think I'd seen the DiDonato and Jarnagin reference before.

A PDF of the paper is freely available. Maybe you can also take a look at a "classic" implementation: Algorithm AS 63 by Majumder and Bhattacharjee, published 1973 in "Applied Statistics". The original Fortran code can be found in various places, at the moment I am doing some tests in VBA with a slighty modified version. This code seems to require more iterations, but each does just a few operations.

However, IMHO the continued fraction approach posted earlier in this thred seems to work just as well. In both methods the crucial point seems to be the threshold where the switch between the calculation of I_x(a,b) and I_1-x(b,a) occurs.

Dieter