Interesting ways to generate a random normal deviate

[Dason]

I was thinking of ways to generate a random normal observation from a single random uniform(0,1) observation. I have at least 3 interesting ways in my head and 1 dull way to do this. I'll share some of the more interesting ways later. The dull way is to use Inverse Transform Sampling to directly transform the uniform into a normal.

Any other interesting ideas? They don't need to be implementable in practice necessarily and you can assume you have a truly random uniform deviate with exact precision.

[chiro]

Hey Dason.

In the Bayesian MCMC, you have a few methods that deal with pseudo-random sample generation. Some that involve the uniform include the following:

http://en.wikipedia.org/wiki/Metropo...ings_algorithm

http://en.wikipedia.org/wiki/Rejection_sampling

http://en.wikipedia.org/wiki/Gibbs_sampling