Relative performance of different exposure modeling approaches for sulfur dioxide concentrations in the air in rural western Canada

Table 1 Comparison to alloyed gold standard constructed as a mean of observed measurements from a given location in 2002 when there were at least two measurements (2 to 24; average = 6, N = 666).

Exposure Assessment Method for annual mean in 2002			ρ_TO ^a	MSE^b
Type of method/model	Model description^Nomenclature	Use of measurements^c
No model	one measurement per location^M1	2002	0.67	0.15
Distance-decay	contextual data only^M2	None	0.21	0.28
Regression	Effects used in prediction
	fixed & random, BLUP^M3	2001	0.77	0.03
	fixed effects^M3f	2001	0.33	0.12
Bayesian	Prior
	lognormal^M4	2001 & 2002	0.68	0.15
	normal mixture from regression model M3^M5	2001 & 2002	0.74	0.09
	normal mixture from context, M2^M6	2002	0.28	0.30

a: correlation with alloyed gold standard; b: mean squared error; c: only one measurement per location in 2002

ISSN: 1471-2288