Table 3 Correlations and normalised root mean squared error (NRMSE)* between personal measurements and exposure estimates (same pollutant, except absorbance as noted) for all subjects and subset with > and ⩽65% of time spent at home
Personal measurements (log-transformed)compared to modelled concentrationsPearson’s r correlations (NRMSE %)
NO, n = 127NO2, n = 127ABS, n = 120PM2.5‡, n = 124
LUR home§0.49 (7.6%)†0.18 (13%)†−0.11 (18%)0.07 (17)
LUR home+work0.55 (6.9%)†0.28 (13%)†−0.10 (18%)0.10 (17)
Ambient monitors IDW monthly0.54 (10%)†0.05 (13%)0.29†† (18%)†0.12 (17)
Subset analysis with >65% of total
sampling session at home (n = 61)
    LUR home0.72 (15%)†0.26 (15%)†−0.19 (24%)0.30 (20%)†
    LUR home+work0.72 (10%)†0.26 (15%)†−0.14 (24%)0.29 (20%)¶
    Ambient monitors IDW monthly0.59 (13%)†0.06 (15%)0.34 (22%)†0.10 (21%)
Subset analysis with ⩽65% of total
sampling session at home (n = 67)
    LUR home0.18 (14)†0.14 (20)†−0.07 (20)−0.07 (16)
    LUR home+work0.31 (14)†0.31 (20)†−0.08 (20)−0.08 (16)
    Ambient monitors IDW monthly0.49 (12)†0.04 (20)0.26 (19)†0.12 (16)

  • *The normalised RMSE is the root mean squared error divided by the range of the personal measurements. The RMSE represents the difference between values predicted by the exposure estimates and the observed values (personal measurements, in this case) and is defined as the square root of the mean squared error (measured units); the NRMSE is expressed as a percentage.

  • †Indicates that the Pearson r correlation is significant at p<0.05; all others non-significant.

  • ‡Personal measurements as PM2.2 were compared to ambient and LUR data measured as PM2.5.

  • §Land-use regression values for NO and PM2.5 are monthly averages, whereas absorbance and NO2 are annual averages.

  • ¶Personal absorbance was compared to ambient monitored PM2.5, since absorbance is not measured by the ambient monitoring network.

  • ABS, absorbance; IDW, inverse distance weighted; LUR, land-use regression.