COMPARISON OF THE GENERAL LOGARITHMIC FORMS’ POWER OF STATISTICAL CRITERIA OF THE HARMONIC MEAN USING THE HYPOTHESIS OF NORMAL DISTRIBUTION OF SMALL SAMPLE DATA
Background. The purpose of the article is to compare the probabilities of type I errors for the statistical chi-square test and two new statistical tests with the harmonic mean in normal and logarithmic forms. Materials and methods. It is proposed to use three statistical criteria when making a...
Main Author: | |
---|---|
Format: | Article |
Language: | English |
Published: |
Penza State University Publishing House
2020-12-01
|
Series: | Известия высших учебных заведений. Поволжский регион:Технические науки |
Subjects: |
Summary: | Background. The purpose of the article is to compare the probabilities of type I
errors for the statistical chi-square test and two new statistical tests with the harmonic
mean in normal and logarithmic forms.
Materials and methods. It is proposed to use three statistical criteria when making a
decision. It is proposed to solve the problem of different scales of three different criteria
by replacing each criterion with an equivalent neuron with a binary quantifier. The
quantizers are tuned to give equal probabilities of type I and II errors.
Conclusions. It is shown that the considered group of artificial neurons has significant
prospects for practical application, since it has an extremely low correlation
coupling. |
---|---|
ISSN: | 2072-3059 |