UDC 330.43 (075.8)
DOI: 10.36871/2618-9976.2023.07.003
Authors
Irina V. Orlova,
Candidate of Economic Sciences, Professor, Financial University under the Government of the Russian
Federation, Moscow, Russia
Abstract
The work is devoted to the development of linear regression models under conditions of multicollinearity of regressors. In this case, the modifications of the orthogonalization method proposed in this article are used, based on the fact that in paired regression, the residual from the regression is not correlated with the dependent variable. The methods under consideration make it possible to reduce the degree of multicollinearity of regressors, and to predict the values of an endogenous variable, use the most appropriate regressors from a content point of view, regardless of the degree of their correlation.
Keywords
Linear regression, Spatial variables, Multicollinearity, Orthogonalization of variables