Estimation of Coefficients in Systems of Regression Models
Alexander P. Sarychev · Journal of Automation and Information Sciences · 2003
A problem of coefficient estimation in a system of regression models is considered. In the studied system, observation errors of output variables of the modelled object are statistically interdependent and output variables can be determined, generally speaking, by different regressor sets. It is demonstrated, that the fact of statistical interdependence can be used for more accurate restoring values of object outputs, which are not observed. A quality functional of regression equation system, which is a multidimensional analogue of J-optimality of regression model, is introduced. An iterative calculation scheme for coefficients of a regression model system, which minimize the introduced functional, is suggested.