Fig 1.
Middle 95% of measurement error distribution (truncated at ± 5 μg/m3) as function of true PM2.5, positive correlation.
Fig 2.
Middle 95% of measurement error distribution (truncated at ± 5 μg/m3) as function of true PM2.5, negative correlation.
Fig 3.
Supralinear C-R functions.
Fig 4.
Examples of C-R functions (PM2.5 composition simulations, % toxic (City 1/City 100).
Fig 5.
The relative slope test for supralinearity.
Table 1.
Detecting true supralinearity with splines in a cox proportional hazards model.
Table 2.
PM2.5 composition effects on supralinearity.
Table 3.
Non-PM2.5 risk factor effects on supralinearity.
Fig 6.
Effect of measurement error on change in slope ratio compared to no measurement error, positive correlation case.
Fig 7.
Effect of measurement error on change in slope ratio compared to no measurement error, negative correlation case.