The Research On Regression Estimation Method Of GDP Accounting

Gross Domestic Product (GDP) is an important indicator to measure a country's comprehensive strength, the research, establishenment of GDP model is meaningful and significant.This paper is dividing into two parts:The first part is to introduce the method to estabilish mathematical model.Firstly, we introduce the linear regression model, which consist of general linear regression, GDP regression based on gray theory, forward regression, principal component regression, robust regression.Then introduced the theory of non-parametric regression and lead to the semi-parametric regression model. The paper set out the estimation procedures about the least squares estimator of the semi-parametic regression model.Finally, introduce some hypothesis testing of the linear regression model and regression models of data rearrangement.The second part is empirical analysis of indicator data.GDP is influenced by various economic factors,so this paper selected twelve economic indicators which is highly related with GDP and have strong reliability. Firstly, we establish the general linear regression, GDP regression based on gray theory, forward regression and principal component regression based on the data. Secondly,many scholars at home and abroad doubt the reality of China's statistics, so this paper use the robust estimation to establish the robust model.Thirdly,considering that GDP is highly correlated with the time T, we add the unknown function to reflect the factors that can't be linear impact on GDP,constructed semi-parametric regression.Oly five models passed the hypothesis test. By comparing the fitness, robustness and predictive of the model, we concluded that the semi-parametric has the best fitness and predictive.It is can be used as an effective regression method to estimate GDP.The robust regression for outliers has a strong anti-interference, which can reduce the impact of the outlier of the sample.At last, I use the rearrangement method to establish the GDP regression of the thiry-one provinces.