Normal distribution on 32E and 21S

[Paul Dale]

I'd expect something like Newton's method for Q^-1. Calculating the CDF gives the PDF at the same time, so the first derivative is free. Likely, there will be an attempt to get a decent initial estimate and possibly some bounding or correction applied to avoid shooting off somewhere where it shouldn't be.

A higher order method is also viable. Second and higher order derivatives also come for free. Thus, Halley's method is also possible. This quickly becomes a balancing game between code size and speed of convergence.


Pauli