The Method for Linear Regression Models Constructing Based on the Sharing of Measured Data and Expert Assessments
Владимир Ефимович ГВОЗДЕВ, Oksana Bezhaeva, Dinara Akhmetova, Alexander Levkov · 2019
The paper discusses the classical approaches to the construction of regression models on the basis of independent and dependent random variables distribution laws.Regression dependencies are one of the main tools for constructing empirical descriptive models, at present moment the theoretical apparatus for constructing regression models has been developed.In the paper analysis of approaches to the construction of the one-dimensional regression dependencies is carried out The analysis allows to conclude that the known methods for constructing of regression models focused on the processing of jointly observed measured data.The authors of the paper propose the method for constructing linear based on the sharing of measured data and expert assessments.Methodological basis for ensuring the comparability of measured data and expert analysis is to convert it to a form of random variables distribution laws.Transformation of expert assessments to the form of continuous random variables distribution laws allows to developed the formal procedure for the construction of linear regression based on the sharing of measured data and expert assessments.Due to the proposed method for constructing of linear regression dependencies in the paper is shown an example of the assessments calculating of paired correlation coefficient and parameters of linear regression dependencies Keywords-regression models, jointly observed measured data, the sharing of measured data and expert assessments, random variables distribution law, linear regression dependencies, paired correlation coefficient I.In [3], a method for constructing non-parametric onedimensional regression dependencies is described, based on 21st