Again, this consistent mortality analysis approach across source apportionments allowed a direct comparison of the daily mortality effect estimates across the various source apportionment analyses in each city.
Keeping the health effects model consistent across the various source apportionment researchers and methods allowed a separate discernment of the extent to which variability in the source apportionment step contributed to variability in the ultimate health effects analyses results.
As noted in Table 1, research groups from seven institutions, using various source apportionment approaches, participated in this workshop.
The goal was to evaluate the consistency of the various source apportionment methods in assessing source contributions to daily PM2.5 mass mortality associations.
As described in Hopke et al. (in press), the various source apportionment analyses from each of the participating research groups were inter-compared in two ways: a) by comparing the mass contributions attributed to each of the sources; and b) by calculating the correlation coefficients between the source contributions from PM2.5 from the various groups within source groups.
In addition, as the various source apportionment results were to be employed as inputs into a daily time-series mortality analyses, the time-series intercorrelations of their respective daily estimates of source impacts were also evaluated and intercompared across source categories in each city.
Based on the relative sizes of these intergroup intercorrelations for each of the source types in these two cities, the soil-, sulfate-, residual oil-, and salt-associated mass components were generally seen to be most unambiguously identified by the various source apportionment methods, while vegetative burning and traffic were less well correlated across groups.
To this end, the various source apportionments' resulting mean mass contributions and estimated percent excess deaths per 5th- to 95th-percentile increment increase by source-apportioned PM2.5 were intercompared and then analyzed (within each city) by analysis of variance (ANOVA) and a GLM.
Instead, the same pre-existing reference PM mass and constituent data sets from two cities (Washington, DC, and Phoenix, AZ) were sent to various leading source apportionment research groups in advance of the workshop (in December 2002), and each group individually analyzed the same data sets for daily source PM2.5 contributions.
Further, receptor modeling through multi-linear regression of the absolute principal component scores (APCS-MLR) provided apportionment of various sources/factors in respective regions contributing to the river pollution.
