Partially Linear Autoregression Models And Its Application

Some estimates and application of partially linear time series models have been studied, which includes Kernel estimation and partial kernel smoothing estimate of partially linear autoregression models and these applied analysis. The contents of the thesis are as follows:1. AR linear regression model is studied in this paper, which includes the classification of the choice of methods in estimation based on AR models, and compared with several methods. Then, the model is applied to China's coal production forecast monthly analysis, results showed that: Since the linear regression model is relatively simple, not very well reflected in the practical factors, so it will cause to a bigger forecasting error .2. Firstly, this paper have done some reaserach on the selection of significant variables,determining the optimal bandwidth and the estimation of parameters and non-parametric on partially linear autoregression models. Fitting with the minimum mean square error of thinking in the semi- parametric Regression Model significant variables identified method, and combination of linear regression model of the band, ultimately determine the parameters and non-parameters of the partially linear autoregression models.3. Partial kernel smoothing estimate and its deviation in partially linear autoregression model have studied, and the windows parameter is choiced by use of generalized cross-validation function. Finally, the partially linear autoregression model based on corporate goods price indices of coal,oil and electricity is established, and the better results proved that our methods to be an effective estimate.