EPA interim final report on WTC background study, April 2003

NYC 9/11 Public Portal Document

As with any statistical evaluation, decisions made regarding the data must be taken into consideration when discussing and applying the results. For this evaluation, there were several decisions that were made regarding the data sets that affect the confidence in the estimated background concentrations.

The first is that when a sample was reported to be below the detection limit, it was assigned a value of Y? the detection limit. A sample that is listed as being below the detection limit may correspond to a concentration that is actually or virtually zero, or it may correspond to a concentration that is very close to the detection limit. By utilizing 56 of the detection limit for samples that are below the detection limit, the true background concentration may be over- or under-estimated. However, this approach consistently assigns a mid-point value that should result in an estimation that is closer to the actual mean than if either zero or the detection limit had been assigned for each sample that was below the detection limit. This is an important concept given that many of the data sets evaluated contained a large percentage of samples that were below the detection limit. As illustrated in Table 8-3, the UCLs presented in Table 8-1 would decrease or increase, respectively, by approximately the same margin if either zero or the detection limit would have been substituted for the samples that were below the detection limit. Data sets with a frequency of detection of less than eight percent would vary by more than 50%, although using 56 of the detection limit provides an estimate value that is within the middle of the range and should be a better estimation of the true mean.

Another decision that was made was to calculate UCLs for all of the data sets regardless of the sample size or frequency of detection. EPA generally estimates UCLs for data sets that have a sample size of greater than or equal to ten. For this assessment several UCL calculations were performed on data sets with a sample size of less than 10. The resulting UCLs for these data sets would be viewed with less certainty than UCLs calculated from larger data sets, but as described above, we were unable to achieve the sample size we desired for every sample type or compound. Additionally, there were multiple data sets for which all of the samples were reported to be below the detection limit. Some of these data sets had varying detection limits which made choosing a single value problematic. To circumvent this problem, data sets that had variability due to varying detection limits also had UCLs estimated, even if the frequency of detection was 0%. These estimated UCLs were identified as theoretical values because the mean that was calculated, from which the UCL was derived, was based only upon the detection limit and it does not include actual detected concentrations. The resulting theoretical UCL for these data sets is generally near the highest detection limit that was used.

Conclusions The evaluation of the data collected from this Study was able to:

provide estimates of baseline levels orbackground concentrations of compounds that were identified as COPCs related to the World Trade Center collapse.

WORLD TRADE CENTER BACKGROUND STUDY 41

NYC-WTC_000145916