| Model specification is an important part of econometric modeling process;they are also an important research direction of econometric theory.The essence of econometrics is to describe the economic phenomenon by data,and finally to reflect the change of economic data and to express in the form of model.In whole research framework of econometric model,model specification is the first place in the process of modeling.When the model is specified correctly,the estimation and test results obtained by the model can be applied to the analysis of economic data.Only econometric model reflect the operation law of economic system,it can provide a solid foundation for economic policy evaluation and economic system structure analysis.When the model is specified by mistake,then any estimation and test methods can not compensate for the error in the model structure.At this point,the results from model can not accurately describe the characteristics of the changes in economic data,and are likely to be wrong,which will bring great risks to economic research.Therefore,it is necessary to study the problem of the model specification;it has important theoretical significance and application value.Model specification is a discussion about the choice of variables,the form of the regression functions and the distribution of residuals when establishing econometric models.Here the mainly research problem is to determine the form of regression function.Semi-parametric regression model is a kind of important econometric model;it not only has the advantages of both parametric and nonparametric regression model,but also can reduce the dependence of the parameter regression model on the assumptions of the data.Due to the advantages of semi-parametric regression model effectively alleviate the dimension trap problem of the nonparametric regression model,semi-parametric regression model is widely used in various fields of social life.So,research of semi-parametric regression model from the perspective of theory and application has crucial significance.Similar to the parameter regression model,semi-parametric regression model also has the possibility of specification by mistake.As the theory of semi-parametric regression model is not perfect and has widely application prospects,so the model specification of semi-parameter regression model will act as the key point in the paper.At the beginning,the paper systematically reviews and summarizes the domestic and foreign literatures about model specification of nonparametric and semi-parametric regression model,fully understands the research status of model specification at home and abroad,and focus on the study of the specification of semi-parametric regression model,which provides a theoretical support for the topic of the article.Based on the basic framework of model specification research of nonparametric regression model proposed by Yatchew(2003),model specification problem of partially linear models and single index models in semi-parametric regression models are also incorporated into the research framework.In fact,researches on those problems are eventually attributed to regression tests on the residuals.The study extends the research scope of model specification and provides a new research method for specification problem of semi-parametric regression model.Moreover through in-depth study,the paper simplifies the steps of various specification approach,and provides a powerful tool for the researchers who use the tests.In the empirical research,the research results of the paper are applied to the study between financial indicators and stock prices of listed companies in China.Through model specification approach,it obtains the suitable econometric model to analyze the relationship between financial indicators and stock prices of listed companies in China,provides theoretical basis for the research of this aspect.The innovation of the article is mainly reflected in the following three points:first,reviews and summarizes the domestic and foreign literatures of nonparametric and semi-parametric regression model specification.According to the different method of research,the adaptability to various methods are discussed,making specification theory of nonparametric and semi-parametric regression models more advanced,providing a simple method for researchers who use those models.Secondly,the specification approach of partially linear models and single index models in semi-parametric regression models are improved,taking test process include in the general research framework,further improving research methods for specification of semi-parametric regression model.Finally,the research results of the article are applied to the study between financial indicators and stock prices of listed companies in China,we analyze them by using the semi-parameter model and get a good conclusion.On this basis,the results are compared and analyzed with these from the parameter linear model usually used in the empirical study.Our empirical results show that the fitting result of the semi-parameter model is better than that of the parameter and nonparameter model.The research not only provides theoretical basis for modeling between financial indicators and stock prices of listed companies in China,it also provides empirical evidence for the semi-parametric model specification. |
