الفهرس 
SHPTER 1Only 14 pages are availabe for public view
 |  | from | 167 |  |  |
| | | from | 167 | | |
Abstract
The use of moving average time series model has played an important role in the analysis of time series. Such models have proved to be useful to researches in many fields. The identification step plays an important and difficult part in time series analysis because other steps depend on it. Therefore many Bayesian and non- Bayesian identification techniques were developed.
The major classical or non-Bayesian tool used for identification of moving average time series models is Box-Jenkins (1976) methodology. This methodology depends on comparing both the autocorrelation function (ACF) and the partial autocorrelation function (PACF) with the theoretical ACF and PACF, which have a characteristic pattern to determine the order of the models. This comparison makes the identification subjective.
In time series modeling, it is desired to have a parsimonious model. Therefore this study uses time series models where some of the coefficients are set equal to zero. Such models are referred to as subset of moving average (MA) time series models.
The most important advantage in the new parameterization is the efficiency with which subset of MA models with large q (q is the maximum order of MA model) may be identified. This advantage is important for long and nonlinear time series which are becoming available in massive dataset being collected in a variety of scientific fields.