Application Of Composite Quantile Regression For Linear Time Series

The time series model, one of the mathematical model, is widely used. The most important node of the time series analysis is parameter estimation. The traditional method of parameter estimation is the least square method for the time series model. But, as we all know, the actual data is difficult to satisfy the hypothesis of LS method. Quantile regression provides an effective solution to the problem. Composite quantile regression is a general form of quantile regression. It not only inherits many advantages of the quantile regression, but also gives better results for parameter estimation and model prediction.In this paper, we use the composite quantile regression as subject investigated, and study it’s estimation effect for linear time series:ARIMA model. The main work is as follows:1. We show the CQR’s expression of stepwise estimation for ARIMA model.2. We compare CQR method to the LS method and QR method through numer-ical simulation for ARIMA model.3. By empirical analysis, we study the prediction effect of CQR method for ARIMA model.