?
Верификация эконометрической модели с учетом априорных ограничений на структурные параметры
The article describes a method for verification of a statistical model, which is, firstly, the time series is represented by original data and, secondly, is linear in the estimated parameters. Experience in statistical calculations on real empirical data shows that the most well-known and conventionally used in the practice of econometric modeling of mathematical-statistical methods (least squares, maximum likelihood method, and similar methods) often do not ensure successful verification theoretically required forms of econometric models. The developed method is called an alternative method of linear regression (AMLR–method) provides an account of a priori restrictions on the absolute values and signs of the parameters identified by the model. The AMLR based on the concept of best linear index, known in the theory of statistics from the end of the 1950s. Mathematically AMLR it based on the method of principal components. The conditions of application AMLR method in econometric modeling and methods of transformation of the initial statistical information to ensure correct application of the developed evaluation procedures. Special problems of the proposed method is to determine the level of accuracy of approximation of the dependent variable of the model. In this regard, to assess the level of precision of the statistical model verifiable using the AMLR, developed an original method of decomposition of the time series on the regular and stochastic components. The properties of the proposed method of decomposition analyzed and given a numerical illustration of its use in econometric calculations.