Evaluation Of Data Quality In Chinese Provinces Based On Robust Regression

Accurate statistics of GDP is an important basis and guarantee that a country or region makes correct macroeconomic policies. The quality of statistics not only determines the prestige of China’s statistics department, but also has a virtual significance to China’s international status. In addition, due to the mature of robust regression theory and statistical diagnosis technology, evaluating data quality from outliners diagnosis has become an important development direction. Therefore, it has theoretical and practical significance to discuss the outliners diagnosis based on robust regression and to choose statistical model in economics to assess the quality of GDP data.This paper mainly evaluates the regional GDP data quality from the point of outliners diagnosis. This thesis compares four types of robust regression by monte carlo simulation, and then gets the better regression which is less affected by outliner in small sample size. After diagnosing and analyzing the GDP data of31areas in China with the method of robust regression, this paper draws the conclusion of data quality assessment. This paper contains four parts:the first part mainly summarizes the connotation of statistical data quality and evaluation methods, and focuses on the application of robust regression and outliner diagnosis in GDP data quality assessment; the second part compares four robust regression with the method of stochastic simulation, getting better robust regression which is less affected by outliners in small sample;the third part detects the outliners in China’s31areas and compares the results with OLS method; the fourth part evaluates the outliners of31areas in2011with robust principle regression to confirm the results of third part, and calculate the confidence interval of regional GDP. The research results that the robust LTS regression is less affected by outliners of different types and ratio under small sample; the percentage of outliners in most area is under10%and focus onf economic transformation or economic crisis; the GDP growth of some areas is over estimated in2011.The innovation of the paper is using stochastic simulation technology to discuss the advantages and applicability of robust regression method under small sample, and apply the robust LTS to evaluate comprehensive diagnosis and analysis of historical data. The study in this article is not only beneficial to the robust regression theory, but also makes some further discussion for data quality assessment methods.