Generation of normal distributions revisited

Takayuki Umeda

DOI: https://doi.org/10.1007/s00180-024-01468-3

IF: 1.4049

2024-02-24

Computational Statistics

Abstract:Normally distributed random numbers are commonly used in scientific computing in various fields. It is important to generate a set of random numbers as close to a normal distribution as possible for reducing initial fluctuations. Two types of samples from a uniform distribution are examined as source samples for inverse transform sampling methods. Three types of inverse transform sampling methods with new approximations of inverse cumulative distribution functions are also discussed for converting uniformly distributed source samples to normally distributed samples.

statistics & probability

What problem does this paper attempt to address?

The paper primarily explores how to generate random samples from a standard normal distribution and proposes new inverse transform sampling methods for 1-dimensional and 3-dimensional standard normal distributions. Specifically, the paper addresses the following key issues:

1. **Importance of generating normally distributed random numbers**: In numerical simulations of statistical physics, such as plasma particle simulations and Monte Carlo simulations, the initial particle distribution significantly impacts the final results. Therefore, it is necessary to generate a set of random numbers that are as close to a normal distribution as possible to reduce initial fluctuations.

2. **Evaluating different types of uniform distribution samples as source samples for inverse transform sampling**: The paper examines two types of uniform distribution samples as starting points for generating normally distributed random numbers and discusses three inverse transform sampling methods based on newly approximated cumulative distribution functions.

3. **Proposing new inverse cumulative distribution functions for 1-dimensional and 3-dimensional normal distributions**: To convert uniform distribution samples into normal distribution samples, the paper proposes a new, analytically invertible approximation method for the inverse cumulative distribution function. This includes approximations for the 1-dimensional standard normal distribution (i.e., the inverse of the error function) and the 3-dimensional standard normal distribution.

4. **Comparing different inverse transform sampling methods**: Through numerical tests, the paper compares the proposed inverse transform sampling methods with the classical Box-Muller transform method. The results show that the new methods can generate random samples more efficiently while ensuring accuracy.

In summary, the main goal of the paper is to improve and optimize methods for generating multi-dimensional normally distributed random samples, particularly for applications in high-performance computing languages such as C and Fortran. These methods aim to enhance computational efficiency and reduce bias, thereby improving the quality of normally distributed random number generation in scientific computations.