Table 3.
Model | Variable | βa | SEb | IQR × βc | p-value | VIFd | Global statistics |
---|---|---|---|---|---|---|---|
PM2.5 (μg/m3) | Intercept | 46.04 | 1.70 | <0.001 | NA | Adjusted R2 = 0.65 | |
Length of first-class road (3000m buffer), m2 | 1.09 × 10−4 | 2.03 × 10−5 | 3.74 | <0.001 | 1.280 | RMSE = 3.12 μg/m3 | |
Farmland area (2000m buffer), m2 | 7.68 × 10−7 | 2.04 × 10−8 | 4.37 | 0.001 | 1.163 | LOOCV R2 = 0.56 | |
Water area (5000m buffer), m2 | −1.04 × 10−6 | 2.01 × 10−7 | 0.38 | <0.001 | 1.731 | ||
Distance to the nearest forth-class road, m | −7.19 × 10−3 | 2.94 × 10−3 | −1.72 | 0.02 | 1.321 | ||
BC (μg/m3) | Intercept | 6.09 | 0.30 | <0.001 | NA | Adjusted R2 = 0.78 | |
Length of first-class road (500m buffer), m | 2.96 × 10−4 | 6.09 × 10−5 | 0.52 | <0.001 | 1.189 | RMSE = 0.51 μg/m3 | |
Length of third-class road (100m buffer), m | 1.08 × 10−3 | 3.83 × 10−4 | −0.38 | 0.010 | 1.126 | LOOCV R2 = 0.70 | |
Water area (4000m buffer), m2 | −4.40 × 10−7 | 6.52 × 10−8 | −0.14 | <0.001 | 1.200 | ||
Industrial area (4000m buffer), m2 | 4.44 × 10−6 | 9.83 × 10−7 | 0.17 | <0.001 | 1.130 | ||
Farmland area (2000m buffer), m2 | 2.17 × 10−7 | 3.49 × 10−8 | 0.90 | <0.001 | 1.114 | ||
NO2 (μg/m3) | Intercept | 22.39 | 0.94 | <0.001 | NA | Adjusted R2 = 0.73 | |
Length of first-class road (700m buffer), m | 8.78 × 10−4 | 1.48 × 10−4 | 2.46 | <0.001 | 1.287 | RMSE = 3.02 μg/m3 | |
Length of third-class road (100m buffer), m | 4.67 × 10−3 | 1.38 × 10−3 | 1.77 | 0.002 | 1.115 | LOOCV R2 = 0.66 | |
Distance to the nearest first-class road, m | −1.25 × 10−3 | 2.75 × 10−4 | −1.67 | <0.001 | 1.381 | ||
Industrial area (5000m buffer), m2 | 1.25 × 10−5 | 2.54 × 10−6 | 1.96 | <0.001 | 1.296 | ||
Emission of NOx from traffic (5000m buffer), t | 1.60 × 10−2 | 6.21 × 10−3 | 1.90 | 0.014 | 1.164 |
AbbreviationsPM2.5, particulate matter with aerodynamic diameter≤2.5 μm; NO2, nitrogen dioxide; BC, black carbon.
β is the regression coefficient of each predictor variable.
SE is the abbreviation of standard error estimate of each predictor variable.
IQR is the abbreviation of the interquartile range of each predictor variable, the contribution of each variable to predicted concentrations (β × inter-quartile range (IQR)).
VIF is the abbreviation of Variance Inflation Factor.