any algorithm for computing the inverse normal cumulative distribution function?

bnubtm · February 20th, 2009, 5:04 pm

I use MT for uniform random number. I don't want to use box-muller for standard normal and try to find a good algorithm for computing the inverse normal cumulative distribution function. any suggestion? Thanks a lot.

AVt · February 21st, 2009, 12:35 pm

I guess you want a random generator for normal distributed numbers - then you better search for that (instead of inverting), Google should help you to find "Ziggurat" by Marsaglia.

gjk77 · February 21st, 2009, 3:23 pm

Shaw, W.T., 2007, Refinement of the Normal Quantile, Simple improvements to the Beasley-Springer-Moro method of simulating the Normal Distribution, and a comparison with Acklam's method and Wichuras AS241

Athletico · February 21st, 2009, 3:47 pm

I agree with AvT, this might help: http://www.doornik.com/research/ziggurat.pdfBut if you insist on using an inverse cumnorm function applied to uniform variates, check: An algorithm for computing the inverse normal cumulative distribution function

bnubtm · February 23rd, 2009, 5:41 pm

Thank you all. Actually I was looking for a generator which has a better tail distribution. I used Acklam's method before. Is Ziggurat better or not? Thanks.

helix · February 24th, 2009, 6:49 pm

Have you seen Moro's approximation. Much better than Akman. I believe there is a Wilmott article on this somewhere.