Based On Time Sequence Control Chart Performance Evaluation And Application Of The Model

As we know, the volatility is an inevitable phenomenon in the development of economy. Drastic economic fluctuations are not only the threat to the stability of economic development and the quality of economic growth, but also the negative impact to people’s living level. Based on time series model, this thesis studies deeply the effective monitoring and early warning to economic process by the application of statistical process control technology. The main results of the thesis are as follows:1. Based on the known 2-order auto-regression process model, the algorithm of average run length (ARL) of two of Shewhart type control charts are studied. Discrete finite Markov chain imbedding and integral equation methods are respectively applied to calculate ARLs of modified Shewhart control chart and Shewhart residual control chart for 2-order auto-regression process. The performance of both charts is compared through the numerical results and some suggestions for application and selection of charts are provided. Through the empirical study on the time series data of Gross Domestic Product (GDP), the specific chart is chosen and applied to monitoring the process which supplies scientific control method for macroeconomic process and promotes the spread and application of control charts technology.2. For the 2-order auto-regression process with unknown parameters, further study of the effect of estimated values of parameters to the performance of control charts are studied. In fact, the effect of the process parameter to the performance of control charts is obvious and the time series process needs to be estimated, so the influence of parameters’ valuation to the performance of control chart is necessary to be analyzed. The study improves the practical value of the application of control charts. Some conclusions and the transformation of the performance of control charts which arises from parameters’ valuation are summarized by numerical results.3. For the clustering property of fluctuations in economic or financial data, the process control method based on the conditional heteroscedastic model is provided. The combination of the regressive model and time series model is chosen to fit the actual data series and the generalized auto-regressive conditional heteroscedastic (GARCH) model is used to describe the volatility clustering. Based on the models, the modified control chart on the residual square series and the modified EWMA residual control chart are designed. The former shows the worse performance than the latter and better performance than GARCH type control chart by means of an empirical example.4. For one economic system including several non-stationary time series variables, the residual series is stationary when the co-integration relationship among these non-stationary variables is existent. So the process control of the total system can be realized at the same time control charts are carried to residual series. This approach avoids effectively not only the "pseudo regressive" problem but also the information loss caused by the difference transformation of sequence, and the target of monitoring the total system is realized better.