And Parameter Estimation Method Study Of Granger Causality Model

Brain research has great scientific significance and philosophical significance. Functional brain imaging has played an important role in the brain research.Among them, functional magnetic resonance imaging (fMRI) is a very effective technique study brain function and be widely used in brain function research.Therefore, the fMRI data has become a hot research From studying whether there was connectivity between brain cortexes to the influence of brain cortexes, that is from functional connectivity to effective connection,many methods have been proposed. Commonly used methods of functional connectivity are correlation analysis, coherence analysis, general linear model, Independent component analysis, principal component analysis and cluster analysis.Commonly used methods of effective connection are structural equation modeling, dynamic causal models and Granger causality models. In recent years, Granger causality model is used to examine effective connection of different brain cortexes by many scholars because of its simple calculation.Processing fMRI data with Granger causality model to effective connection of different brain cortexes face a problem that must be solved, that is order selec-tion.Only a certain order, we can better estimate model parameters and residuals, improve validity of Granger causality model. This article first explains the impor-tance of order selection,and then introduced the six information criterion function of order selection and the three estimation method of parameter estimation, meanwhile comparatively analyze the six information criterion function and the three method of parameter estimation by simulation experiment. Finally,the relatively better solu-tion of model order and parameter estimation is given according to the experimental results. In order selection, the paper put aside the past method used to obtain the residual variance, introduce new methods. be used to approximately take the place of the residual variance.