Skip to main content
. Author manuscript; available in PMC: 2021 Feb 1.
Published in final edited form as: Atmos Environ (1994). 2019 Nov 14;222:117130. doi: 10.1016/j.atmosenv.2019.117130

Table 2. Daily cross-validation results: summary statistics for RMSE.

Mean, standard deviation, and other quantile summaries of daily root mean squared error (“RMSE”) for PM2.5 concentration predictions in μg/m3. The methods considered are ordinary least squares (“OLS”), inverse distance weighting (“IDW”), universal Kriging(“UK”), downscaler, random forests (“RF”), support vector regression (“SVR”) and Neural networks (“NN”). Methods use either CMAQ and/or other geographic covariates (“Covs”).

Summary Statistic Mean SD Min Q1 Median Q3 Max
OLS (CMAQ) 4.71 1.68 2.00 3.53 4.32 5.28 12.15
OLS (Covs) 4.57 1.53 1.80 3.48 4.28 5.39 10.15
OLS (CMAQ+Covs) 4.15 1.42 1.64 3.18 3.85 4.76 9.34
IDW 3.25 1.14 1.58 2.42 2.94 3.89 7.87
UK (CMAQ) 3.09 1.12 1.38 2.27 2.81 3.69 7.85
UK (Covs) 3.27 1.44 1.45 2.39 2.99 3.79 18.52
UK (CMAQ+Covs) 3.19 1.17 1.39 2.36 2.87 3.78 7.81
Downscaler(CMAQ) 3.15 1.19 1.39 2.31 2.88 3.73 8.63
RF (CMAQ + Covs) 4.21 1.46 1.95 3.23 3.86 4.80 10.76
SVM (CMAQ + Covs) 3.87 1.53 1.60 2.82 3.53 4.46 10.31
NN (CMAQ + covs) 3.89 1.37 1.65 3.00 3.55 4.46 10.96