基于主成分回归的杭州市PM2.5含量的综合分析  

The Comprehensive Analysis of PM2.5 in Hangzhou Based on Principal Component Regression

作　　者：宋红凤[1] 徐登可[1] 陆丹妮 SONG Hongfeng;XU Dengke;LU Danni(Department of Statistics,Zhejiang A&F University,Hangzhou,Zhejiang 311300,China)

机构地区：[1]浙江农林大学统计系,浙江杭州311300

出　　处：《经济数学》2020年第4期141-147,共7页Journal of Quantitative Economics

基　　金：浙江省高校重大人文社科攻关计划项目(2018QN037)。

摘　　要：为探究杭州市PM2.5含量与其他指标之间的关系,搜集杭州市2019年的AQI指数和AQI 6项指标进行分析.首先通过相关分析揭示PM2.5与其余5项指标的相关性,发现各指标与PM2.5均呈线性关系,又因为其余5项指标之间存在显著的线性相关性,故自变量之间存在多重共线性.为了消除多重共线性的影响,对这5项指标提取主成分,再基于主成分分析做回归分析.最后对模型进行统计诊断,按照一定标准筛选出原始数据中的强影响点,删除强影响点后得到最终模型.To explore the PM2.5 component content in Hangzhou with other indicators,we collect indicators and indexes about AQI and other five indicators.Firstly the correlation analysis reveals that the other five indicators have linear relationship with PM2.5,and they have significant linear relationship with each other,so we eliminate the multicollinerity.Secondly we extract the main components with the five indicators,and do regression analysis based on principal component analysis.Finally we diagnose the model and delete the strong influence spot to get the final model.

关 键 词：PM2.5 相关分析 主成分分析 统计诊断 COOK距离 W-K统计量 

分 类 号：O212[理学—概率论与数理统计]