Inverse of normal distribution


jvalero
05-09-2003, 03:22 AM
Hi,

Given a probability p from 0 to 1, I would like to calculate the INVERSE of the NORMAL (GAUSSIAN) distribution.

How can I do it? Can I use the method described in chapter 7.2, page 289 of "Numerical Recipes in C"? I do not understand pretty well if that algorithm covers what I want to do.

Thanks in advance,

Jose Luis.

Bill Press
05-10-2003, 04:12 PM
Jose Luis,
I assume that you mean the inverse of the cumulative normal distribution, since you say you want to input a probability between 0 and 1.
The cumulative normal distribution is related (with a constant scaling on the independent variable) to the error function, which is discussed at NR page 220.
The routine erfcc is fast enough that for most purposes you can invert it by using the simple method of bisection to solve the equation erf(x)=P for unknown x with known P. See NR page 353.
Good luck!
Cheers,
Bill P.