Environment/Climate Change
Applications of Statistical Analysis Methods to Sparse Data at Low Fine Particulate Matter Concentrations Jacob Kremer* Jacob Kremer Kremer Kremer Kremer Kremer Exponent, Inc.
Fine particulate matter (PM2.5) has been demonstrated to adversely affect human health and mortality. However, considerable debate exists over estimation of the precise Concentration Response Function (CRF) describing the relationship between PM2.5 and all-cause mortality. To evaluate potential improvements to CRF modeling techniques when applied to sparse data at low exposures, Al-Kindi et al., 2019 was chosen as the basis for a replication analysis and method development. Techniques used for exposure data modeling included Inverse Distance Weighting and Kriging (Gaussian Process Modeling); CRF modeling included Poisson GAM and Pohar Perme estimation of net survival. Overall, the replication analysis showed agreement with Al-Kindi et al.’s estimated treatment effects but found a lower mortality hazard at lower levels of PM2.5 exposure, suggesting that mortality is driven by higher levels of exposure.