You are here:
Exploring the added value of imposing an ozone effect monotonicity constraint and of jointly modeling ozone and temperature effects in an epidemiologic study of air pollution and mortality
Citation:
Crooks, J., L. Neas, AND A. Rappold. Exploring the added value of imposing an ozone effect monotonicity constraint and of jointly modeling ozone and temperature effects in an epidemiologic study of air pollution and mortality. Presented at Eastern North American Region International Biometric Society Meeting, March 10 - 13, 2013.
Impact/Purpose:
This abstract will be presented at the Eastern North American Region International Biometric Society (ENAR) Meeting, March 10-13, 2013, Orlando, FL
Description:
Abstract: A number of epidemiologic studies have shown that both ozone and temperature are associated with increased risk for cardio-respiratory mortality and morbidity. However, their joint effects are not characterized as well as their independent effects. Furthermore, the impact of requiring the ozone risk rates to be everywhere positive has also not been fully examined in the nonlinear case. The current study seeks to understand the impact on the inferred risk surface of (1) jointly modeling the nonlinear effects of ozone and temperature versus modeling them independently, and (2) of imposing a positive monotonicity constraint on the ozone risk. To this end, a flexible Bayesian hierarchical model was developed to allow for these modeling choices. The independent, nonlinear effects of temperature and ozone on mortality are modeled using M- and I-spline basis functions, respectively, while outer products of these same functions are used to model the joint effect. Positive monotonicity on the ozone risk rate is imposed by placing a prior distribution on the I-spline and the I-spline/B-spline outer product coefficients that has a steep drop-off below zero. The model hyper-parameter controlling the shape of the drop-off can, however, be chosen to soften or harden the threshold as desired. The model was applied to the data from the US National Morbidity, Mortality, and Air pollution Study for 95 major US urban centers between 1987 and 2000. For comparison, results are examined with respect to those obtained under the assumption of linear effect of ozone without positivity constraint (Bell et al, JAMA 2004). Disclaimer: This work does not reflect official EPA policy]