Robust Regression Method And Its Application Research In Econometric Modeling

In collecting the survey data in practice, the occurrence of error is inevitable. The presence of outliers tends to have a serious impact on the results of the model, even distorts the results of the analysis. In order to get the correct model estimation results, it needs to detect and remove outliers in the data processing. Removal of the outliers is also the normal processing of outliers for traditional statistics. But outliers are not always harmful, and are also contain certain economic, natural or political factors, which should attract our attention. So the traditional outlier processing methods have certain limitations. But modern measurement method-robust regression method modern robust regression can provide unbiased estimation which is not affected by outliers. But the robust regression method is used in chemical, medicine, geography, such as natural science, and relatively rare in the areas of the economy. Statistics plays an important role in economic modeling, but many research institutions and domestic and foreign scholars question the authenticity of China’s economic data. Therefore, it has profound practical significance and theory significance that the robust regression methods are combed and chosen the suitable robust methods to evaluate China’s economic data.In this paper, based on the current situation of the robust regression, some robust regression methods are integrated. The abilities of anti-interference the abnormal value and diagnosis outliers of M(Huber)、M(Biweight)、M(Hampel)、GM、MM、LTS and LAV estimates are simulated though Monte-Carlo simulation technique. And GM、MM、LTS estimates have an advantage. What’s more they are applied to China’s GDP statistics on the reliability of the data quality assessment. We can evaluate authenticity of the GDP data, according to the outliers diagnosis and the robustness of total factor productivity growth rate.The empirical results show that the GDP is relatively reliable.This article mainly includes three parts. The first part focuses on robust regression and the robust diagnostic method. The second part compares OLS and seven kinds of robust regression resistance and the abnormal value diagnosis efficiency, through the simulation of different types of outliers. The third part GDP data quality are assessed by GM、MM、 LTS、OLS methods, and we can learn the steady diagnosis can get more information. What’s more to some extent, robust estimation can replace the difference.The innovation of this article:in statistical diagnosis simulation using diagnosis index of average replace probabilistic MSR index to quantify the degree of abnormal data; robust estimation replace the difference of time series which are not stable; Instead of the subjective judgment method this paper using unit root test to learn that Whether the TFP growth rate is stable; at the same time using the outlier diagnosis and TFP logical evaluation standard to evaluate GDP data quality and longitudinal comparatively analyze on these two methods.