Analysis,Modeling And Application Of Structure-Transition Model Of Time Series

The analysis and modeling of the financial time series is a very important study realm in financial metrology. But nonlinear problem in financial data and nonlinear economic metric model in financial time series is an all new research topic in this realm. Existing study nonlinear time series structure transition model to have two mainstreams, Firstly, one takes time as structural change point. Secondly, threshold autoregression method was advanced by Tong in 1978, which take variable as structural change point. Because of analytical method taken time as turning point have to be subjective to affirm time of the structure change order, in the method not more objective. So gain the conclusion is quit differet. While the significant expositive variety in the model take place significant variety continuously in a short time in addition, the traditional time series model can't examine the structural change of the model probably. Comparatively, analytical method taken variable as turning point can avoid imperfection. Variable can be observed directly, therefore we can use it directly. In context, the author do lots of research on structure transition model of time series. Beside, the author especially discussed the modeling of threshold cointegration system, parameter estimation of threshold vector error correction model, confirming of conversion zone scope and testing of threshold cointegrating.The main work of dissertation include:1. Modeling fashion of threshold cointegration system be carried on to subdivide. People use it to modeling have selectivity more.2. Due to our country reduce the nominal interest rate continuously in recently years, and economy appears slight deflation. The nominal interest rate and the price level appear homology of descend trend, but this kind of idea can not judgment Fisher effect at our country economy show remarkable existence. In context, using the system modeling methed in threshold cointegration system to set up threshold vector error correction model, to reevaluate our country nominal interest rate and inflation rate relationship.3. Hansen and Seo(hereafter HS)suggest a grid search algorithm that allows joint estimation of cointegrating vector and threshold value in a bivariate threshold vector error correction model with a single cointegration vector. However, HS algorithm is difficult for estimating large system with multivariate vector correction model and multiple cointegrating vector.HS algorithm involves a joint grid search over the threshold value and the cointegrating vector. Thus, in context,the author do lots of improve on HS algorithm .So, this makes the HS algorithm flexible for managing large system.