Русский
!

Conference publications

Abstracts

XXII conference

The method of the system of forecasting

Koshkin Y.L., Martynova O.N.

Russia, 610000, G. Kirov, street of Karl Marx 77

1 pp. (accepted)

System of econometric equations is not currently used to solve many problems of the economy. For example, for prediction or simulation to cast their application is not clear.

Attempts to identify existing methods (indirect, two - and three-step OLS and others) lead to obscure models that are clearly not relevant source data. In particular, almost always turns out that the residual sum of squares is more common (as well as the total sum of squares does not depend on the model, the residual does not match the observed data (as was done in [1])). This is because to ensure the so-called identifiability have a priori to reduce the fraction of the variables for some strange rules. In our opinion a priori to exclude anything not necessary, and the source to be taking into account all the variables.

It should also clarify the concepts of endogenous and exogenous variables. Endogenous is the dependent variables, they are used for prediction by the model. They always stand to the left of the equal sign. All variables on the right - exogenous, that is, independent variables that are determined outside the system and affect the endogenous variables, but not depend on them. Thus, equally marked (and has a certain practical sense) variable can be in the system and endo - and exogenous. However, the econometric sense (in their roles) are different. When identification system (search parameters) values of exogenous variables are taken from the sample, and to predict offered to take them from other models.

Let as other models use time series (if all variables can be predicted as a function of time). These projections are preliminary and can be substituted in the system of equations that build in the usual way by OLS. A priori, all variables depend on all, but after selection factors leave only statistically significant.

As additional models can also be used to preliminary expert forecasts, scenario predestination. The proposed two-step procedure allows one to obtain the system forecasts in different fields: Economics, physics, engineering.

Literature. 1. Eliseev, I. I., Kuryshev S. C., Kasteev So Century. and other Econometrics. - M.: Finance and statistics, 2007. 576 p.



© 2004 Designed by Lyceum of Informational Technologies №1533