Method of Choosing the Best Distribution Law for a Random Variable Based on Experimental Data
I. A. Klyavin1, A. N. Tyrsin2
1Chelyabinsk State University, ul. Br. Kashirinykh 129, Chelyabinsk, 454021 Russia
2ivank@mail.ru
2Reliability and Life of Large Systems and Machines Scientific-Engineering Center, ul. Studencheskaya 54a, Ekaterinburg, 620049 Russia
at2001@yandex.ru
Keywords: random variable, distribution law, probability density, random sampling, Monte Carlo statistical simulation, goodness-of-fit test
Pages: 18-25
Abstract
This paper describes a new method for choosing a distribution law from a given set of experimental data that provides the best fit to a measured random variable. The method is based on a comparison of empirical distributions constructed for the original sample with the set of given laws using a continuous mapping of the distribution function onto the interval [0, 1]. As a result, the distribution for which the corresponding value of the functional is maximal is taken as the most probable distribution law for the original sample. Examples of implementing the method using statistical Monte Carlo simulation are given
|
