Statistical tests of the IBM PC pseudorandom number generator.

Area of Science:

• Biomedical research
• Computer science
• Statistics

Background:

• IBM personal computers are prevalent in biomedical research and medical practice.
• Many applications utilize the BASIC programming language and model stochastic processes.
• These applications often rely on the Microsoft random number generator.

Purpose of the Study:

• To evaluate the performance of the Microsoft random number generator.
• To identify specific seeds that produce non-random number sequences.
• To recommend reliable seeds for use in scientific applications.

Main Methods:

• Statistical analysis of random number sequences generated by the Microsoft generator.
• Testing for uniformity and serial correlation using established statistical methods.
• Evaluation of various seeds to determine their suitability.

Main Results:

• Evidence of non-uniformity in generated number sequences for certain seeds.
• Demonstration of serial correlation within sequences for problematic seeds.
• Identification of acceptable seeds that satisfy randomness criteria.

Conclusions:

• The Microsoft random number generator exhibits significant flaws with specific seeds.
• Careful selection of seeds is crucial for reliable stochastic modeling in research.
• Recommended seeds ensure the integrity of random number generation for scientific computing.