Research On The Evolution And Application Of Linear Causality Modeling Method

Causal relationships are prevalent in many studies,such as height affects weight,corporate image affects business performance,class atmosphere affects achievement and so on.Causality research has been a hot topic of empirical research.In the causality,the linear causality is the main research object.The linear causality modeling method has also become a common statistical tool to deal with the causal relationship.The methods mainly include linear regression analysis,path analysis,structural equation model,Hierarchical linear model.In the study of linear causality,the choice of cause is the key.Social studies often find that there are parallel relationships between some reasons,such as investment and consumption;some have progressive relationships,such as income and consumption,income will affect consumption if they are all the reason.There are also some reasons have nested relationship,such as class atmosphere and student learning attitude,which are the two levels of reason.Because the relationship between the reasons is different,causality can be divided into three categories: parallel,progressive,nested.Based on the causal relationship type,the linear causality modeling method is also evolved along two main lines.One is based on the parallel relationship to the progressive relationship,from the linear regression analysis to the path analysis,then to the structural equation model.One is based on the parallel relationship To nested relationship,from linear regression analysis to multi-level linear models.As a commonly used statistical tool,linear causality modeling method is often studied and applied independently.The systematic research on its evolution is lacking.The research in this paper is necessary and practical.First of all,introduce the theory of the each stage of linear causality modeling methods,include the basic ideas,basic forms,parameter estimation,testing,Advantages and disadvantages of various methods.Secondly,the application research is carried out through three cases.One is the analysis ofinfluencing factors of GDP,based on linear regression,path analysis and Hierarchical linear model.The other is consumer trust analysis based on structural equation model.Then,the characteristics of various methods and evolution law are summarized.Finally,the further development and application of linear causality modeling method are prospected.It is found that the linear causality modeling method is a rigorous methodological system,which is not a simple denial of the former method,but a process of development and innovation in the inheritance.From the perspective of solving the problem,the first main line,from the linear regression analysis to the structural equation model,the middle variable and latent variable problms are solved,also allow the independent variable have error.For the second line,from the linear regression analysis to Hierarchical linear model,the nested data problm is solved.At the same time,in the evolution of linear causality modeling method,some assumptions are relaxed,the parameter estimation method is more scientific and reasonable,and the application of sample information is more sufficient.Linear causality modeling method is more and more reasonable,and its applicability is more and more extensive.This paper studies the linear causality modeling method from the perspective of methodology,and then sums up its evolution law,which is novel in related research and can provide reference for its application and follow-up development.