基于指示变量对城市空气数据真实性判别分析

Discriminant Analysis of Urban Air Data Authenticity Based on Indicator Variables

目的以AQI为衡量标准的城市空气数据,建立新的分析方法对数据真实性进行判别。方法以北京市为研究对象,搜集得到1月10日前后20d的AQI指数与5种污染物浓度数据,即细颗粒物(PM2.5)、可吸入颗粒物(PM10)、二氧化硫(SO2)、二氧化氮(NO2)、一氧化碳(CO),借助SPSS软件分别进行回归分析,得到2个回归方程式,然后建立基于指示变量的多元回归比较模型,对2个方程式偏回归系数和常数项进行比较。结果 2个空气数据回归方程的偏回归系数检验结果为:F_1=0.05,F0.05(5.8)=3.69,F_1<F0.05(df1,df2),偏回归系数无显著性差异;常数项的检验结果为:F_2=0.89,F0.05(5.12)=3.11,F_2<F0.05(df1,df2),常数项也无显著性差异。可以看出北京1月1日到1月20日的空气质量数据不存在造假情况。结论造成空气质量数据产生不真实性的原因主要有2个:主观性的数据统计错误和非主观性的数据统计错误。因此可以从改革政府考核方面的硬性要求、建立支持环境大数据发展的组织结构和改善空气质量这3个方面来避免城市空气数据造假情况。

Objective For the urban air based on the AQI,a new analysis method was established to discriminate the data authenticity.Methods By taking Beijing as the object,the data of AQI and five kinds of pollutant concentration data before and after January 10 th were collected.Five kinds of pollutant included PM2.5,PM10,SO2,NO2 and CO.So two regression equations were obtained with the help of SPSS software.The regression model was established based on indicator variables.Then the authenticity of AQI was determined.By comparing the partial regression coefficient and the constant term of the two equations,we could judge the authenticity of AQI.Results The test results of the coefficients of two partial regression equations were,F_1=0.05,F0.05（5.8）=3.69,F1 〈 F0.05（df1,df2）.There was no significant difference between the partial regression coefficient.The test results of the constant term of two regression equations were F_2 =0.89,F0.05（5.12）=3.11,F2〈 F0.05（df1,df2）.There was no significant difference between the constant term of the two equations.So the air quality data from January 1 st to January 20 th in Beijing were true.Conclusion There are two main reasons why the air quality data are not authentic：subjective data errors and non-subjective data statistical errors.Therefore,to avoid the fraud of city air data,reforming rigid requirements of government assessment,establishing the organizational structure to support the development of big data environment,and improving the air quality.