Presented by: 
Bhaswati Ganguli and Sugata Sen Roy, University of Calcutta
Fri 11 Oct, 12:00 pm - 1:00 pm

The NMMAPS and APHENA initiatives are well known efforts to model the effect of short term particulate matter on daily mortality in North America and Europe respectively.  In 2002, the Health Effects Institute, Boston organised its PAPA initiative to replicate similar studies across several sites in Asia. The initial objective was to use a common protocol for analysis based on semiparametric Poisson models regressing daily mortality on city-wide average exposures adjusted for meteorological confounders. This talk presents the challenges which arose in modelling the Indian dataset which was characterised by a high percentage of missing exposures and spatial differences in exposure. An initial adhoc approach is discussed followed by two alternative approaches based first on latent variable modelling  and second on an application of  the EM algorithm for modelling an autoregressive series of counts.