The paper discusses the problem of approximation accuracy for selected probability distributions (Student’s t, χ, geometric, Bernoulli, Poisson) using limit distributions used, among others, in statistical and reliability research, as well as the conditions for their applicability.1. INTRODUCTION
The use of probabilistic methods is common not only in engineering.
In applications of probability theory, as well as in probability theory and statistics themselves, one of the fundamental problems in their application is the ability to effectively calculate the probabilities of random events and determine the quantiles of probability distributions (random variables).
The cumulative distribution function and, for continuous distributions, the probability density function are used for this purpose. However, the distributions of the variables under consideration are often unknown (a typical situation in estimation problems and statistical hypothesis testing).
Unfortunately, even knowing the functional form of the cumulative distribution function (density) does not guarantee the ability to effectively calculate the sought values – computational calculations for exact distributions are complicated. In such situations, exact distributions (known or unknown) are approximated by typical probability distributions for which calculations are easier or which have been tabulated. These are most often the so-called limit distributions. For their effective application, it is necessary to know the error we make by replacing the approximated distribution with the approximating distribution. (…)
The article contains 27058 characters. < /p>
Source: Logistyka Magazine
Approximation of certain probability distributions – PDF file for download








