Theories Research And Application Of Two Questions In Multivariate Statistical Analysis

Multivariate Statistical Analysis is an embranchment of Mathematical Statistics which developed so rapid these 30 years,researching the dependable statistical regularity among multivariate variables of objective things.Every individual should have many observation data,and the method of analyzing these numerical data is called Multivariate Statistical Analysis.Since computer is quite widespread in the world,kinds of statistical software are advanced constantly;theories of Multivariate Statistical Analysis are developing endlessly;methods of Multivariate Statistical Analysis are widely used in every field.This article is mainly discussing the sample statistical quantity to do Empirical Orthogonal Function(EOF),Canonical Correlation Analysis(CCA)in Multivariate Statistical Analysis and related nonlinear time series' concepts and basic models.Ordinary method of empirical orthogonal function only extracts empirical space orthogonal function.The Second Chapter of this article presents a new method: adapting the empirical space orthogonal function with time weight function,and analyzing the empirical time orthogonal function and space weight function can extract more information.The problem of NLCCA is familiar but not easy to solve.At present,NN method used in solving the model of NLCCA is complicated in programming and could not get the expression among the variables.The Third Chapter of this article talks about the forecasting of rainfall around the lower reached of Yangtze River,and advances a polynomial model of NLCCA and its two solving methods:Stepwise Selecting and Principal Component Method.using these two new methods we can find the nonlinear CCA forecasting factors successfully.