ESTIMATION OF POLYNOMIAL COEFFICIENTS IN PARAMETRIC GMDH ALGORITHMS BY THE IMPROVED METHOD OF INSTRUMENTAL VARIABLES

Methods of estimation, of coefficients by GMDH and by the method of instrumental variables (MTV) are compared. Numerical modelling has shown that the error of identification of the estimates of linear regression coefficients depends on the value of the conditionality index of the normal equations X(T)X. It is shown that MTV has an advantage under considerable dispersions of the interference as well as in the case when the law of the interference distribution differs from the normal. MTV algorithm is improved by means of a decrease of the value of shifts achieved by a linear, approximation of the data sampling.