空气状况影响着人们的生活与健康，人们越来越重视空气质量情况。本文先对近几年北京市空气质量情况进行描述性分析，发现2012年到2013年北京市空气质量大幅下降，之后有小幅度回升，空气质量整体呈夏季好冬季差的特征，不同区域差异明显。通过数据模拟，比较了一般最小二乘回归和3种稳健回归方法（M估计、LTS回归和MM估计），得到LTS估计对异常值相对更稳健，因而采用LTS方法对北京市空气质量情况进行回归分析。从年度空气质量变化来看，从2003年到2015年，北京市空气质量达到及好于二级的天数与北京的人口数、城市绿化覆盖率和降水量正相关，与北京的GDP和机动车拥有量负相关；从2016年每日空气质量情况来看，北京市空气质量指数与监测点位置所在经度成正相关，与所在天数、监测点位置所在纬度和当天最高气温成负相关。可以从以上几个方面对北京的空气进行合理地治理。

Air quality affects people's lives and health, has attracted more and more attention. Based on the descriptive analysis of Beijing's air quality in recent years, Beijing's air quality was greatly decreased from 2012 to 2013 and then had a small rebound. The air quality is good in summer and poor in winter. The air quality in different areas is obviously different. Through data simulation, this paper compares the general least squares regression and three robust regressions (M estimation, LTS regression and MM estimation). The LTS estimates are the most robust against abnormal values, so the LTS estimate is used to analyze Beijing's air quality. From the point of the annual air quality, from 2003 to 2015, the number of days of Beijing air quality reached and better than two is positively correlated with the population of Beijing, urban green coverage and precipitation, is negatively correlated with the GDP and motor vehicle. From the daily air quality in 2016, Beijing air quality is positively correlated with longitude of monitoring point, is negatively correlated with the number of days, latitude of monitoring point and the highest temperature.