Statistical Analysis Of Nonlinear Regime Switching Models And Its Applied Research

In recent years, many economists study results showed that the time series data of many macroeconomic indicators (such as GNP, RPI, exchange rates, etc.) showing the characteristics of nonlinear wave, and theclassic linear model has been unable to portray increasingly complex market economic phenomena, thus promoting more widely used of nonlinear models in economic and financial fields. Therefore, this paper extended the three nonlinear time series regime-switching models (TAR, STAR and MSA) and made statistical analysis (mainly the Bayesian analysis) of the models, then used the three models to reveal the regular pattern of the development of China’s macroeconomic phenomenon, which results have stronger economic practical significance.First, the article analyze generalized conditional heteroscedasticity threshold autoregressive (AR-TAR-GARCH model) with Bayesian method, and then select the China’s Retail Price Index (RPI) time series as the object of study, the research results show that the order difference sequence of the RPI exist cluster effect and the impact of negative external shocks to DRPI series is greater than positive external shocks, called leverage effect, our government should interfere appropriatly to prevent price fluctuate by a large margin, it is recommended that throughout the relevant departments to take effective measures appropriate interventions to prevent excessive retail commodity price fluctuations.Secondly, in recent years, the trend of appreciation of RMB excited more and more attention of the import&export traders and government departments, this paper select the Sino-US exchange rate time series as an object of study, through the analysis showed ESTAR-GARCH model can well describe the Sino-US The trend of changes in exchange rates, exchange rate increment has both the symmetrical nonlinear characteristics and cluster effect, the ESTAR-GARCH (1,1) model was used to forecast Sino-US monthly exchange rate during the recent two years, the high forecasting precision demonstrates that the ESTAR-GARCH model is very well and RMB will revalue during the future ten months.Finally, the paper randomly generates an obedient Markov Switching model (MSA) time series by editing the program, and use Bayesian simulation to analyze the sequence, it is estimated that The MSA model parameters, the results suggest the stationary distribution of the Markov chain introduced MSA Bayesian analysis method is feasible.The full text summarized the state of the field of regime-switching models and synthesized previous studies, analyzed the regime-switching autoregression models and their combined models with GARCH based on the modeling idea of regime-switching models’nonlinear and classical&Bayesian statistics method, the models were applied to Chinese macroeconomic research and achieved desired effect.