PM2.5是当前中国面临的首要大气环境问题，本文根据影响PM2.5浓度变化的部分气象因素和空气污染物，采用多元统计方法分析风力、湿度、二氧化硫、二氧化氮、一氧化碳、臭氧这六个因素对PM2.5浓度的影响。首先，分析气象因素和空气污染物与PM2.5的线性相关性和变化趋势，发现风力、臭氧的变化与PM2.5浓度变化呈负相关，其他因素的变化与PM2.5浓度的变化呈正相关；其中二氧化硫、二氧化氮、一氧化碳和PM2.5之间存在较强的线性关系；湿度、风力、臭氧和PM2.5之间的线性关系比较弱。以风力、湿度、二氧化硫、二氧化氮、一氧化碳、臭氧为自变量，以PM2.5浓度为因变量建立回归模型，并对残差进行检验。发现模型的残差的方差是非齐性的，并且残差之间存在相关性，于是对因变量进行了Box-Cox变换解决方差的非齐性问题，对残差建立时间序列模型，最后得到动态回归模型。 利用时间序列模型对PM2.5的浓度进行建模，利用最小二乘方法估计模型参数，根据AIC最小的准则确定最优阶数，得到PM2.5浓度满足的ARMA模型，并进行预测。考虑到PM2.5和其他变量的关系，利用多元时间序列中的向量自回归模型（VAR模型）分别对风力、湿度、PM2.5组成的向量和二氧化硫、二氧化氮、一氧化碳、臭氧、PM2.5组成的向量建立VAR模型，分析变量之间的动态关系，并对PM2.5的浓度做出预测，比较和分析模型的预测结果。

PM2.5 is the primary problem facing the current Chinese atmospheric environment, in this paper, according to the meteorological factors and air pollutants that affect the change of PM2.5 concentration, using multivariate statistical analysis to analyze the effect of wind, humidity, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone on concentration of PM2.5. Firstly, analyze the linear correlation and the trend of meteorological factors and air pollutants and PM2.5, finding that changes in wind, ozone concentration was negatively correlated with PM2.5, positively correlated with changes in other factors and PM2.5 concentration; there is a strong linear relationship between sulfur dioxide, nitrogen dioxide, carbon monoxide and the PM2.5; the linear relationship between humidity, wind, ozone and PM2.5 is weak. With wind, humidity, sulfur dioxide, nitrogen dioxide, carbon monoxide and ozone as independent variables, the PM2.5 concentration as the dependent variable to build the regression model, and to test the residuals. Finding the variance of residuals is inhomogeneous and there is correlation among the residuals, the dependent variables of Box-Cox transform to solve the problem of non homogeneity of variance, establish the time series model of residual error and finally get the dynamic regression model. Establish the time series model of the concentration of PM2.5, parameters of the model are estimated using the least squares method and determine the optimal order according to the criterion of minimum AIC. ARMA model of PM2.5 concentration to meet, using the ARMA model can predict the concentration of PM2.5. Considering the relationship between PM2.5 and other variables, establish vector autoregressive model (VAR model) of wind, humidity and PM2.5 vector and sulfur dioxide, nitrogen dioxide, carbon monoxide, ozone and PM2.5 vector separately, analyzing the dynamic relationship between the variables, predicting the concentration of PM2.5 and comparing the prediction of different models.