Air pollution modelling for birth cohorts: a time-space regression model

Table 2 Final model for the urban area

Variables	Percentile			Estimate per IQR^a	95 % CI lower	95 % CI upper	Cumulative Adj. R²
Variables	25	50	75	Estimate per IQR^a	95 % CI lower	95 % CI upper	Cumulative Adj. R²
Sqrt (vehicles in 100 m buffer) * season^b	1728	3696	6117	−0.219	−0.265	−0.172	0.291
Log (NO₂ from dispersion model) log(NO ₂ concentration)	3.21	3.28	3.37	0.052	0.039	0.065	0.341
Log (NO₂ from AQM Payerne) log(NO ₂ concentration)	2.3	2.68	3.03	0.216	0.181	0.252	0.372
Sqrt (vehicles in 100 m buffer)	1391	1997	3074	0.404	0.362	0.446	0.437
Log(1/distance to the nearest major road) log(1/m)	−4.08	−2.95	−2.61	0.163	0.144	0.181	0.470
Linear time trend year	2002.6	2005.2	2007.7	0.477	0.387	0.567	0.488
Season (summer = 1, mid-season = 2, winter = 3)^b	1	2	3	0.191	0.118	0.264	0.499
Industrial land use in 300 m buffer percent area	0	0	0.237	0.436	0.384	0.487	0.506
Population in 100 m buffer N	0.95	141	323	0.118	0.097	0.139	0.514
(Total length of major roads in 100 m buffer)^2 (m^2)	26931	48969	147510	0.296	0.259	0.334	0.519
Total length of major roads in 100 m buffer m	164	221	384	−0.414	−0.472	−0.356	0.534
Linear time trend ^2 (year^2)	2002.6²	2005.2²	2007.7²	−0.462	−0.563	−0.36	0.540
Temperature Celsius	3.4	9.05	15.59	−0.081	−0.126	−0.035	0.540
(Boundary layer height)^2 (m^2)	16723	79082	359729	−0.013	−0.024	−0.002	0.541
Total length of major roads in 100 m buffer * temperature	0	1485	3807	0.034	0	0.069	0.541

Model developed without an intercept term. The R² is not provided in the regression output when the intercept is suppressed; we thus manually calculated the R2. The predictors are ordered per decreasing relevance on the basis of incremental R². Most p-values were <0.001; p-value for “Total length of major roads in 100 m buffer * temperature” was <0.05
* indicates multiplication of variables
^aFor land use data (high and low density residential land use and industrial land use) we report the estimate per increase from 0 to 100 % of used area instead of per increase of IQR because data distribution is skewed and IQR would be 0
^bSeason categorised as 1: summer (May to August), 2: mid-season (March, April, September, October), 3: winter (November to February)

ISSN: 1476-069X