Figure 3. RSS relative percent difference between TBP and Ct using PCDD/F congener- and homologue-specific BSAFs and between TBP and Ct calculated using a single overall BSAF

ERDC TN-DOER-R5

September 2004

Figure 3.

RSS relative percent difference between TBP and Ct using PCDD/F congener- and

homologue-specific BSAFs and between TBP and Ct calculated using a single overall BSAF

(formula in Table 4)

Bootstrap Uncertainty. Bootstrap calculations of TBP and uncertainty generally mirrored

those of the RSS method, although with narrower uncertainty intervals (Figure 4). The bootstrap

intervals are nonparametric estimates of statistical confidence intervals based solely on

resampling of input parameter replicates. RSS uncertainty intervals, on the other hand, combine a

method error component with incremental (propagated) error derived from parametric

confidence limits for each input parameter, and thus incorporate more sources of uncertainty,

leading to generally broader uncertainty intervals. Bootstrap mean TBPs determined using

congener/homologue mean BSAFs (Table 6) reflected a positive bias of about 8 percent

compared with the calculated TBPs. The positive bias is indicative of a skewed distribution

resulting from simulations involving ratios.