Generating Random Variables from Distributions

 

Generating Random Variables from Distributions

General Methods for Continuous Distributions

1. Transformation : only suitable for the distribution function F(x) whose inverse F^(-1) is not so difficult to evaluate. The first step is to generate U from the uniform distribution; then solve the nonlinear equation F(x) - U = 0 for x. the solution X has the distribution F.
2. Acceptance/Rejection : invented by von Neumann in the early days of computing, is one of the most useful general tools in random number generation.
3. Ratio of Uniforms : The Ratio of Uniforms method  is essentially a particular form of A/R where the tails of distributions are made more manageable, often leading to short and simple algorithms for many distributions, including discete ones. The key is a simple result by Kinderman and Monahan(1977).

General Methods for Discrete Distributions

1. Discrete Inverse
2. Acceptance/Rejection
3. Ratio of Uniforms
4. Walker's Alias Method

Reference
1. Numerical Methods of Statistics by John F. Monahan Cambridge University Press; 1st edition (February 15, 2001)
2. http://en.wikipedia.org/wiki/Random_generator
3. Matlab Online Help