The Application Of Quantile Regression In Time Series

Time series analysis is one of the modern statistical methods in the pro-cessing of data. Its theory and method have been widely applied to engineeringtechnology, meteorology, hydrology, earthquake, biomedicine, economic man-agement, military science and so on. Time series analysis model the time-ordered data, analyze its dynamic structure and law of development, thenpredict the future.Time series data involves so many areas, and its structure is also com-plicated and diversified. Random error distribution in model will appearsdi?erent features for di?erent data types, such as asymmetric or thick tail.There is not any requirement to the distribution of random errors, when weestimate the parameters in model with quantile regression. Quantile regres-sion could also estimate the quantile function of conditional distribution, anddescribe the features in di?erent locations by estimating function in di?erentquantiles.In this paper, we model the time series data by quantile regression. It'son the basis of studying quantile regression and time series theory, we collectmonthly sales data about Australian wine and analyze this time series datawith quantile regression.Then, we estimate and test the model parametersin di?erent quantiles, further forecast the sales of wine. Therefore, quantileregression can get more complete information compared to the usual timeseries models.