F distribution on the 41C

[Dieter]

(12-17-2015 03:36 PM)quantalume Wrote:  

(12-17-2015 02:56 PM)Ángel Martin Wrote:  I's suggest you look at Jean-Marc Baillard's program collection, which has a very comprehensive section on Statistical Distributions posted at:
http://hp41programs.yolasite.com/statistic.php

Thank you, Ángel. Unfortunately, my office network has flagged that site as containing malware, and I will have to wait until I arrive home later today to visit it. I have seen Jean-Marc's excellent work in the past.

JMB also uses the incomplete regularized Beta function for the F-distribution. Which indeed is the obvious choice here.

I now got a short program for the HP41 running. It uses one unified algorithm for all cases (odd and even d.o.f. in any combination) via the regularized Beta function. Here the (complete) Beta function is evaluated directly, which can be done without much effort because the underlying Gamma code only has to handle integers and half-integers so that the 41's FACT function can be used. This makes Beta reasonably fast and accurate, only both d.o.f. must not exceed 69.

At the moment the incomplete Beta function uses code from the HP67/97 library, which of course can be optimized. That's why I do not want to post this more or less experimental code now. Just want to say it's not that complicated (about 160 lines now). ;-)

Dieter