基于Sieve Bootstrap方法的季节时间序列模型检验问题研究

研究目标：完善季节时间序列模型建模理论，解决建模过程烦琐、各类检验方法的结论差异大以及模型误设定问题。研究方法：基于对各季节时间序列模型的数理分析及比较，提出合理的模型检验程序；再运用Sieve Bootstrap方法，给出季节性单位根检验及确定性季节过程检验的统计量的临界值，并比较基于Sieve Bootstrap的检验方法与HEGY检验、BT检验的异同。研究发现：本文提出的检验程序能有效识别模型，检验统计量有限样本性质优良；实证分析表明，本文提出的检验程序及方法能更有效地识别中国宏观经济数据中的季节性。研究创新：将Sieve Bootstrap方法应用于季节时间序列的平稳性检验及趋势性检验中。研究价值：提出季节时间序列模型检验程序及检验方法，促进其在季节性经济数据中的应用。

Research on the Model Testing of Seasonal Time Series Based on Sieve Bootstrap

Research Objectives: Solving the problems in seasonal time series modeling, such as model testing procedure is complicated, result of different model testing methods are inconsistent and model misspecification, in order to improve the theory of seasonal time series modeling.Research Method:Firstly, we propose the model testing procedure of seasonal time series based on mathematical analysis and comparison across different seasonal time series models. Secondly, we propose the critical values of seasonal unit root test statistics and deterministic seasonal process test statistics based on Sieve Bootstrap approach. Finally, we compare the differences across Sieve Bootstrap based test, HEGY test and BT test.Research Findings:The model testing procedure we proposed can test the seasonal times series model effectively, and the statistics based on Sieve Bootstrap has better finite sample properties. According to empirical analysis, the testing procedure and method we proposed can recognize the seasonality of China''s quarterly macroeconomic variables better.Research Innovation:Applies Sieve Bootstrap to stationary test and trend test of seasonal time series.Research Value:Proposes the seasonal time series modeling test procedure and method, and promotes the application of them in seasonal economics data