基于稳健回归的我国各地区GDP数据质量评估

发布时间：2018-02-28 03:02

本文关键词： 稳健回归异常点诊断数据质量　出处：《天津财经大学》2013年硕士论文　论文类型：学位论文

【摘要】：准确的GDP统计数据是一个国家或地区正确制定宏观经济政策的重要保证和依据。统计数据质量的好坏不仅决定着我国统计部门的威信,而且对中国的国际地位也有重要意义。此外,随着稳健回归理论和统计诊断技术的日益成熟,从异常点诊断角度评估数据质量成为一个重要发展方向。因此,在稳健回归方法基础上讨论异常点诊断技术,选择同时具备统计意义和经济意义的模型来评估各地区GDP数据质量,极具理论价值和现实意义。论文主要从异常点诊断的角度评估我国各地区的GDP数据质量。具体采用蒙特卡洛模拟比较研究了四类稳健回归方法,得到小样本量情况下受异常点影响较小的回归方法,进而在该稳健回归的基础上对我国31个地区的GDP数据进行全面的诊断和分析,最后得出各地区历年的GDP数据质量评估结果。文章主要分为四个部分：第一部分主要概括了统计数据质量的内涵和评估方法,重点综述了稳健回归和异常点诊断方法在GDP数据质量评估中的应用；第二部分采用随机模拟技术比较研究了四种稳健回归估计,得出小样本下受异常点影响较小的稳健回归方法；第三部分用稳健LTS回归方法实证诊断我国31个地区GDP数据中包含的异常点情况,并和基于OLS的异常点诊断结果进行对比；第四部分在异常点诊断的基础上,使用稳健主成份回归方法印证上一章的诊断结果,并计算出2011年我国各地区GDP增长率的置信区间。研究结果表明,在小样本情况下,不同比率和类型的异常点对LTS估计影响较小；1978-2011年大部分地区GDP数据异常点在10%以下,且集中出现在经济转型或经济危机时期；部分地区2011年的GDP增长率存在高估的现象。论文的创新之处在于,运用随机模拟技术讨论了小样本下异常点类型和比率不同时,稳健回归方法的优劣和适用性,并在稳健LTS基础上对我国各地区历年GDP数据做了全面的异常点诊断和评估。本文不仅对稳健回归方法做了有益的比较,更重要的是对数据质量评估方法做了进一步的探讨。
[Abstract]:Accurate GDP statistical data is an important guarantee and basis for a country or region to correctly formulate macroeconomic policies. The quality of statistical data not only determines the prestige of statistical departments in our country. In addition, with the development of robust regression theory and statistical diagnostic technology, evaluating data quality from outlier diagnosis has become an important development direction. On the basis of robust regression method, this paper discusses the technique of anomaly diagnosis, and selects a model with both statistical and economic significance to evaluate the quality of GDP data in various regions, which is of great theoretical and practical significance. In this paper, the quality of GDP data in different regions of China is evaluated from the point of outlier diagnosis. Four kinds of robust regression methods are studied by Monte Carlo simulation. On the basis of the robust regression, the GDP data of 31 regions in China are diagnosed and analyzed comprehensively. Finally, the evaluation results of GDP data quality in various regions are obtained. The paper is divided into four parts: the first part summarizes the connotation and evaluation methods of statistical data quality. The application of robust regression and outlier diagnosis in GDP data quality assessment is reviewed. In the second part, four kinds of robust regression estimation are compared with each other by using stochastic simulation techniques. In the third part, we use robust LTS regression method to diagnose the abnormal points in 31 regions of China, and compare the results with the results of abnormal points based on OLS. 4th based on the diagnosis of outliers, the robust principal component regression method is used to verify the diagnostic results in the previous chapter, and the confidence interval of GDP growth rate in each region of China in 2011 is calculated. The results show that, in the case of small samples, From 1978 to 2011, the outliers of GDP data in most regions were below 10%, and concentrated in the period of economic transition or economic crisis. In 2011, the GDP growth rate in some regions was overestimated. The innovation of this paper lies in the use of stochastic simulation technology to discuss the advantages and disadvantages and applicability of robust regression method with different types and ratios of outliers in small samples. On the basis of robust LTS, this paper makes a comprehensive diagnosis and evaluation of the outliers of GDP data in various regions of China. This paper not only makes a beneficial comparison to the robust regression method, but also makes a further discussion on the evaluation method of data quality.
【学位授予单位】：天津财经大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F224;F127

【参考文献】