Article 4118

Title of the article

TESTING ESTIMATES ON THE BASIS OF INTEGRAL AND BAYESIAN APPROACHES 

Authors

Mikhailov Viktor Sergeevich, lead engineer, Central Research Institute of Chemistry and Mechanics named after D. I. Mendeleev (115487, 16а Nagatinskaya street, Moscow, Russia), Mvs1956@list.ru

Index UDK

621.382.029.6

DOI

10.21685/2307-4205-2018-1-4

Abstract

Background. The aim of this paper is to construct a rule for choosing an effective evaluation based on integral and Bayesian approaches based on the results of tests conducted in accordance with the NВτ-type plan and finding effective estimates in accordance with this rule.
Methods. To find an effective estimate, we used the integral numerical characteristics of the accuracy of the estimate, namely, the total square of the displacement of the expected realization of a certain valuation variant from all possible values of the estimated characteristic from the different values of the parameter of the Poisson distribution law that characterizes the failure flow of the set of products under test.
Results and conclusions. In accordance with the logic of the constructed rule for the choice of effective estimates based on the integral and Bayesian approaches, for tests conducted in accordance with a plan of type NBτ and based on a priori known uniform distribution law , it is possible to sum up: 1) the Bayesian estimate  Qθ loses its effectiveness in comparison with the proposed estimates; 2) caution should be exercised when using the Bayesian approach to the evaluation of test results. When checking the results obtained, it is also necessary to use effective integral estimates. As such, it is necessary to use the T01 and f ; 3) if the tester decides not to use the Bayesian estimate and not take into account a priori a known uniform probability distribution, then the traditional unbiased estimate should be used as the FBI estimate, and in the case of trouble-free tests, should be used the estimate  = e (-g/T04

Key words

Poisson distribution law; exponential distribution; test plan; point estimation; Bayesian estimation

Download PDF

 

Дата создания: 14.05.2018 14:06
Дата обновления: 14.05.2018 15:07