Post Reply 
[WP34s] Regularized incomplete Beta function
05-05-2014, 08:14 PM
Post: #26
RE: [WP34s] Regularized incomplete Beta function
(05-05-2014 11:15 AM)Paul Dale Wrote:  The normal QF estimate used in your Student's-T is slightly different to the one I used in the normal.wp34s file. Only on the < .23 branch in the code.
Is the difference significant?

No. I just simplified the last term 0,254/u to 1/4u. The original version even is a tiny bit better. So you should leave it the way it is. I remember it was also fine-tuned to make sure the Normal quantile safely returns 32...34 digits in just two iterations.

If all the distributions get reworked we could also switch to a simple rational approximation for the Normal estimate. I suppose a (2;2) type, i.e. with five real constants, should be sufficiently accurate. Since the initial guesses for the other distributions rely on a Normal guess we just have do decide now whether to switch or not. Designing such an approximation would require some time. On the other hand, everything also works fine with the current solution.

Dieter
Find all posts by this user
Quote this message in a reply
Post Reply 


Messages In This Thread
RE: [WP34s] Regularized incomplete Beta function - Dieter - 05-05-2014 08:14 PM



User(s) browsing this thread: 2 Guest(s)