METHOD FOR RESOLUTION OF ALTERNATIVENESS IN ESTIMATES OF PARAMETERS OF REGRESSION MODELS

Abstract

Background. Regression analysis is a very effective means of mathematical modeling of complex systems of various nature and scale, allowing you to identify both explicit and hidden trends in their functioning and development. The purpose of the study is to solve the problem of calculating a compromise estimate of the parameters of a linear regression model that arises when it is built by several alternative identification methods. Materials and methods. To achieve this goal, a mathematical apparatus for solving linear programming problems was used. Results. In the general case, the formulated problem is reduced to a computationally complex problem of nonlinear programming. Under some assumptions simplifying the original formulation, the original problem was reduced to a linear programming problem. At the same time, the method of concessions, popular in the theory of decision making, was used. Conclusions. The algorithm described in the paper makes it possible to avoid alternativeness when estimating the parameters of regression models.

Keywords

• regression model
• loss function
• parameter estimation methods
• linear programming problem
• alternativeness
• concession method