A Comparison of Methods for Generating Normal Deviates on Digital Computers

ABSTRACT

Two methods recently developed for generating normal deviates within a computer are reviewed along with earlier proposals. A comparison of the various methods for application on an IBM 704 is given. The new direct method gives higher accuracy than previous methods of comparable speed. The detailed inverse technique proposed yields accuracy comparable with, or better than, most previous proposals using about one-quarter the computing time.

REFERENCES

• 1: G. E. P. Box ASp M. E. MULLER, A note on the generation of normM deviates. Ann. Math. Stat. 28 (1958), 610-611.
• 2: JA~ES W. BUTLER, M:achine sampling from given probability distributions. Symposiuvar in Monte Carlo Methods, H. A. Meyer, ed., Wiley & Sons, New York, 1956, pp. 249-264.
• 3: H. C~M~R, Mathematical Methods of Statistics. Princeton University Press, Princeton, N. J., 1046.
• 4: GEORGE E. FORSYTItE, Generation and testing of random digits at the National Bureau of Standards, Los Angeles. Mon~e Carlo Method, National Bureau of Standards AMS 12 (U. S. Government Printing Office, Washington, D. C., 1951)~ pp. 34-35.
• 5: Cecil Hastings, Approximations for Digital Computers, Princeton University Press, Princeton, NJ, 1955
• 6: D. L. JOHNSON, Generating and testing pseudo random numbers on the IBM Type 701. Math. Tables Aids Comp. 10 (1956), 8-13.
• 7: M. L. JUNCOSA, Random number generation on the BRL high-speed computing machines. Ballistic Research Laboratories, Aberdeen Proving Ground, Md., Report 855 0953).
• 8: H~RM~N KAHN, Applications of Monte Carlo. RAND Report No. RM 1237 AEC, (1954).
• 9: D. H. L~H~.~, Mathematical methods in large scale computing units. Proceedings of Second Symposium on Large-Scale Digital Calculating Machinery, Harvard University Press, Cambridge, Mass., 1951, pp. 141-146.
• 10: D. H. LEI~MS~, Description of "Random number generation on the BRL high-speed computing machines", Math. Rev. 15 (1954), 559.
• 11: Jack Moshman, The Generation of Pseudo-Random Numbers on a Decimal Calculator, Journal of the ACM (JACM), v.1 n.2, p.88-91, April 1954 [doi>10.1145/320772.320775]
• 12: M. E. MULLER, An inverse method for the generation o~ random normal deviates on large-scale computers. Math. Tables Aids Corap. 12 (1958), 167-174.
• 13: J. YON N~U~ANN, Various techniques used in connection with random digits. Monte Carlo Method, Nationa} Bureau of Standards A}VIS 12 (1951), pp. 36-38.
• 14: ~ymposium in Monte Carlo Methods, H. A. Meyer, ed. (Gainesville). John Wiley & Sons, New York, 1956.
• 15: T~ RAND GO~P.: O~e Million Random Digits and 100,000 Normal Deviates. The Free Press, Glencoe, Illinois, 1955.
• 16: O. TxussKr xz~v J. Torn, Generation and testing of pseudo random numbers. ~yraposium in Monte Carlo Methods, H. A. Meyers, ed., John Wiley & Sons, blew York, 1956, pp. 15--27.
• 17: D. TEIcn~oEw, Distribution sampling with high speed computers. Ph.D. Thesis, University of North Carolina, 1953.
• 18: D. F. VOTAW ANX) J. A. RAFFER~V, High speed sampling. Ma~h. Tables Aids Comp. 5 (105~), l-s.
• 19: H. WOLD, Random Normal Deviates, Tracts for Computers No. XXV, Cambridge University Press, 1948.

INDEX TERMS

Primary Classification:
  G. Mathematics of Computing
  G.3 PROBABILITY AND STATISTICS
      Subjects: Random number generation

Additional Classification:
  F. Theory of Computation
  F.1 COMPUTATION BY ABSTRACT DEVICES
      F.1.2 Modes of Computation
          Subjects: Probabilistic computation

General Terms:
Measurement, Performance, Reliability, Theory