A Preliminary Study On The Systematic Risks Of Enterprises In The Third Board Market Discussion

Bayesian statistics is a discipline which devoted to solving computational problems.In Bayes,the Markov chain Monte Carlo(MCMC)calculation is the preferred method for solving a class of computational problems.The in-depth study of the MCMC algorithm is helpful for us to have a deeper understanding of the Bayes principle,which makes us better grasp the use of Bayes in scientific research and life.There are a large number of scholars has in-depth reserch and derivation on the MCMC algorithm in foreign countries,but for the for the MCMC algorithm in the calculation of how specific methods are still relatively small.In the country,the use of MCMC principles and the use of computer simulation literature even less.This study borrows the principle of MCMC method,and uses the concrete realization method of Gibbs sampling,and take advantage of the ability of computer to have random simulation,which can perfectly simulate the ability of normal distribution form and the advantages of ensuring the automation of calculation.Using matlab programming software,how to program MCMC algorithm automation to do a preliminary combination of theory and practice to explore.This experiment we study the poor liquidity of the assets of the beta and other parameters,that is to say the company's yield data for the discrete value,non-continuous.So we need to use the MCMC principle to obtain the company's relevant parameters of the posterior distribution,and get the company's related parameters,then to understand the company's market-related or irrelevant situation.By combining theory with practice,we first summarize the Gibbs sampling algorithm with a basic process: in Bayesian,we see every constant that we do not know as a random variable and subject to a certain distribution,so what we have to do is to get the posterior distribution of the parameters we are interested in.First,we need to assume the form of posterior distribution,and the establishment of the relevant model.Then we assume the hypothesis or simulation of the prior distribution and the initial value,and we will perform a Bayesian linear regression on the simulated continuity yield.A sample value of the posterior distribution of each parameter is obtained,which continues the iteration as an initial value until we get the Markov chain to converge.Once we iterations long enough to get a Markov chain,remove the previous initialsimulation stage(Burn-in period).we get the back of the Markov chain value can be regarded as the subsequent Markov chain to derive the mean and variance of the posterior distribution of the parameters we require,and the resulting mean it can be regarded as the value of the parameter which we are interested in.But we will study the problem of non-current assets,so our yield data are discrete values,and before the Gibbs iteration,we have to make use of Kalman to predict the data which we can not observe the theoretical still prediction.So that we can get a priori or experienced a priori value.This will not lead us to the results of the latter will not converge for a long time the situation.