(50g) Normal Distribution

[Dieter]

(12-08-2018 08:15 PM)Albert Chan Wrote:  The extra precision is basically useless.

z-score = (x–µ)/σ, with all 3 variables non-exact.

Are they? For the standard Normal distribution µ is exactly 0 and σ is exactly 1. So there is a real benefit if the calculation is done with the best possible accuracy.

Sure, the general assumption for all numeric calculations, be it on a HP67, HP50g or a contemporary computer, is that the input is exact. So the input is not √2 but 1,41421356237, and not pi but 3,14159265359. That's a limitation that always applies.

But in other cases the input can be exact. And here there improved PDF offers a benefit.

Dieter