Dependence Research About Financial Markets Based On Bayes-copula

Nowadays, Primary Market and Secondary market both are not only thecore that regulators and investors concern, but also the fields where inventorsand fund-raisers, between whom the market analysts serves as a bridge, competeagainst each other. When the market analysts analysis the unique kurtosis andheavy-tailed characteritic makes the previous analysis based on the normaldistribution deficient. In this paper, the writer will use Copula function, thecombination of Clayton,Gumbel and Frank functions to picture the finacialmarket structures on the whole.In this paper, the first few chapters make a simple review of the meaningand nature of Copula function, and illustrate several genel purpose Copulafunction and the correspondence between the parameters and the function ofseveral correlation coefficients. Then it introduces the financial markets timeseries models, including parameters and non-parametric estimation, andhighlights the kernel estimation method. After that it tells the Bayesian theoryand Bayesian calculation method, And using the Bayesian theory to infers thelikelihood function and posterior density function of triplet linear model basedon Mix-Copula-function and experience of Copula. The paper also talks abouthow to put the bayesian theory into the practice of of estimate of coefficient.and linearly fits the experience-Copula-function and Mix-Copula-function.The text takes the Shanghai Composite Index and Shenzhen ComponentIndex from2009to2013daily historical closing prices as a sample touses kernel density estimation to analyze the marignal distribution of thevaribles. and after what has been done above, it uses the Archimedes Copula-function to build four different Copula function for analysising, and Euclideandistance to test them and finally to select the best one. And compare theoutcome of Bayesian and that of Nonlinear pragamming method, as theoutcomes we can get that the outcome on the Bayesian have the effect of theordinary method, or even better, and mix the Copula function to get thecorrelation between the upper and lower.