An alias method for sampling from the normal distribution (Q1825579): Difference between revisions

The authors, well-known for their contributions to the theory and practice of pseudorandom number generation, propose a new ``alias'' technique [see \textit{D. E. Knuth}, The art of computer programming, Vol. 2: Seminumerical algorithms. 2nd ed., Chapter 3.4.1 (1981; Zbl 0477.65002)] for sampling from the normal distribution. The expected number of uniforms required per normal slightly exceeds 1.06, and a comparison on speed with other methods, including the authors' own, the next fastest [Math. Comput. 27, 927-937 (1973; Zbl 0285.65008)] indicates an impressive superiority.