An Efficient Generator of Uniformly Distributed Random Deviates between Zero and One.
Abstract
The author shows that a uniformly distributed random deviate between zero and one can be constructed by combining two independent random variables: (1) a random exponent drawn from a geometric distribution, and, (2) a random mantissa drawn from a uniform distribution. Algorithms for generating these random variables are developed, and a Pascal procedure combining these into a computationally efficient random number generator for micro computers is presented. The resulting generator is computationally efficient as it uses no floating point arithmetic uses on the average only 3.5 random bytes to construct a four byte random deviate. Finally it is shown empirically that the generator has both an unusually long period and excellent statistical properties. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1983
- Accession Number
- ADA132851
Entities
People
- Arne Thesen
Organizations
- University of Wisconsin–Madison