Study Of Structural Equation Models Based On Compositional Data

In the multivariate problem of various research fields of social science,researchers often encounter latent variables that cannot be measured accurately or directly.The traditional statistical methods will have some limitations in dealing with this problem,so the structural equation model is usually used to solve this problem.The research of Structural equation Model including measuring model and structure model,therefore,it can effectively deal with and test the relationship of explicit variable and latent variable(or latent variable and latent variable),and it can be a practical method of multivariate statistical analysis.Compositional data is a kind of special data which often exists in social,economic and educational fields,that is,each vector in the compositional data satisfies the constraints and non-negativity.In this paper,partial least squares regression of structural equation model is applied in compositional data,through the example analysis of the model can get the modeling has a high fitting degree.In the case studying,the relationship between tourism service and tourism development is analyzed firstly,then joining tourism investment to study the relationship among the three variables.Through the analysis of examples shows the model with sign weighted sum has its advantages and disadvantages,therefore,this method is modified to make the method more reasonable and effective.This paper is divided into five chapters:Chapter I:Introduction.This paper introduces the research purpose and meaning,knowledge of Structural equation model is described and the main research status at home and abroad,and points out the content of this articleChapter II:The introduction and application of simplex space.This paper mainly introduces the related knowledge of composition data and the linear regression model of composition data and its properties.Chapter III:PLS regression of structural equation modeling with two latent variables in simplex space.Firstly,the basic structure of the structural equation modeling is briefly introduced.And then the regression modeling with two latent variables and all variables as composition data is described.Chapter IV:PLS regression of structural equation modeling with multiple latent vari-ables in simplex space.This paper mainly introduces the regression modeling of multivariate latent variables and all variables are composition data.Chapter V:Summary and Outlook.The main research contents of this paper are summarized,and the future research direction are pointed out.