Analysis Of The Impact Of The Financial Crisis On The China's Stock Market Based On Parameter And Non-parameter Methods

The development of the economy of each and every country around the world can be reflected upon the developing level of its stock market,which illustrates the intimacy between a country's economy status and its stock market.The importance of a stock market as a measurement of a country's developing economy equals to the importance of a weather forecast to a worker.The insight study of China's stock market is quite relevant to the recognition of the stage in which a country's economy development is among the world,the meaning of which is of significant importance.Using non-parameter method to model financial time series dataset seems of better performance than parameter method when it comes to fitting bias.The non-parameter modeling method is a new research orientation for modern statistic research field,which leads to a new perspective of modeling dataset.There are no hypothesizes towards either estimating function in the model format or parameters.The result should be completely decided by the dataset itself.This quality marks the flexibility of practical usage of non-parameter method.The most frequently used non-parameter methods include kernel estimation,kernel regression estimation and local polynomial regression estimation.Jianqing Fan presented the ideal of double kernel local polynomial regression method in his article published in 1996.This method can solve the problem that conditional density function is mostly calculated indirectly using density functions.This method can calculate the conditional density function directly using this double kernel local polynomial regression estimation.The key point of using non-parameter method in practical calculation is the value of bandwidth.Most recommended method of choosing the appropriate bandwidth value is the cross-validation method.Jianqing Fan presented this modified cross-validation method in his article published in 2004,which is the completion of his double kernel local polynomial regression estimation method.In this article,the main method is based on the double kernel local polynomial regression estimation method.However,the introduced conception of entropy is also used in the study of stock market.In this thesis,firstly the two representative indices are used as the subject.The S&P500 Index works as a comparison.The time period is chosen from 2010 to 2015,in which period the 2008 economic crisis is certainly not included,the method being used here is parameter modeling--auto-regressive conditional heteroscedasticity method.After analyzing the result of modeling which can lead to some significant conclusions.Secondly,the non-parameter method is also used in this article-double kernel local polynomial estimation method when modeling 19 different index datasets.The time period is chosen from 2002 to 2015.Moreover,the time period is divided into three stages:before the 2008 financial crisis,during the 2008 financial crisis,after the 2008 financial crisis,during which period the entropy is calculated respectively for comparison.Based on the calculated entropy,the influence of the 2008 economic crisis is observed with some conclusions suggested from this study.