A FORTRAN IV version of the Marsaglia rectangle-wedge-tail algorithm and IMSL's commercially available Box-Muller polar method of normal deviate pseudo-random number generation were combined with various linear congruential uniform pseudo-random number generations and tested on a 32-bit-word IBM 370/155 computer. Looking at their relative speed and degree of meeting normality as joint performance criteria placed Marsaglia's method very slightly ahead of Box-Muller's. A commercially available IMSL inverse normal subroutine performed least effectively.
Get full access to this article
View all access options for this article.
References
1.
Chen, E. H.A random normal number generator for 32-bit-word computers . Journal of the American Statistical Association. 1971, 66, 400-403.
2.
International Mathematical and Statistical Libraries . The IMSL Library (Vol. 2) (7th ed.). Houston, Tex.: IMSL, 1979.
3.
Kinderman, A. J. and Ramage, J. G.Computer generation of normal random variables. Journal of the American Statistical Association, 1976, 71, 893-896.
4.
Knuth, D. E.The art of computer programming (Vol. 2: Seminumerical algorithms). Reading, Mass.: Addison-Wesley, 1969.
5.
Marsaglia, G. , MacLaren, D., and Bray, T. A.A fast procedure for generating normal random variables. Communications of the ACM, 1964, 7, 4-10.
6.
Siegel, S.Nonparametric statistics for the behavioral sciences, New York: McGraw-Hill, 1956.
