The program for estimation non-elementary linear regressions with two variables using ordinary least squares

Abstract

Objective. The aim of this article is to develop a program for approximate estimation of regression models specified on the basis of the Leontief production function (non-elementary regressions with two variables) and use it for modeling the unemployment rate in the Irkutsk region.Method. Estimation of non-elementary regressions is carried out using ordinary least squares method. To find approximate estimates, we used a previously developed algorithm that involves solving a very laborious computational problem.Result. Based on this algorithm, a special program was developed in the Delphi programming environment. The program provides for work in manual and automatic modes. In manual mode, according to the specified criteria, the estimates of the model parameters, the residual sum of squares, the coefficient of determination, the Student's criterion, Durbin-Watson's criterion and, for each variable, the number of the binary operation components triggerings on the sample, are determined. In automatic mode, the best estimates of non-elementary regression are determined according to the criteria: residual sum of squares, coefficient of determination, the Student’s criterion and Durbin-Watson’s criterion. At the same time, graphs of all the main characteristics are plotted depending on the key parameter of the model. With the help of the developed program, a model of the unemployment rate in the Irkutsk region was construct.Conclusion. The model construct using the developed program turned out to be better than the traditional model of multiple linear regression. The program is universal and can be used to solve specific applied problems of data analysis.