O szacowaniu parametrów modelu zmian w czasie z uwzględnieniem wahań okresowych

Abstract

Among methods of time series analysis present in statistical packages, most of them are smoothing methods. Although they are efficient in many cases, especially in forecasting, they are unable of describing a model of changes over time. Many statistical textbooks present an analytical method of coefficients, which however is not an efficient method of estimation the parameters of the model. Because of two separate stages of the estimation procedure, one can obtain false results, especially in the context of the proper interpretation of the model. Authors concentrate on two modification proposals of the method of coefficients, which seem to be computationally simple and free of typical drawbacks of the method of coefficients. In the paper, authors also present an iterative method, which can be simply applied in Excel, and additionally shows high efficiency. Comparisons of results obtained by the classical and modified methods are carried out on the basis of simulation experiments and some empirical examples.