Accurate Normal Distribution for the HP67/97

[Dieter]

(12-03-2018 12:44 AM)Albert Chan Wrote:  That is a cool formula ! I was expecting something much worse.

Something much worse would not have fit into the limited HP67/97 memory. ;-)

(12-03-2018 12:44 AM)Albert Chan Wrote:  To prove above correction work:
(...)
Sum it all, and ignore O(h⁴) terms, we get:

Ah, thank you very much. I have never tried a proof for that formula myself. But I wonder what that neglected fourth term with t⁴ will look like. Due to the limited hardware precision it will not make much sense, but anyway. I have a faint idea, but you will know better – Albert?

BTW, the tricky part is calculating t in the first place. Especially evaluating Q(x)–p without too much digit cancellation.

Dieter