AG 2007 10 15 - Laserfiche WebLink

Home Browse Search

ATTACHMENT: "Ozone population exposure analysis for selected urban areas". For the record, the suggestions I made for clarification in this section were ignored for the most part. (Obviously they cannot be addressed now, as they are not mere errata.) More specifically: Page 22: No estimate has been given of the effect of using the school children's D and A as a surrogate for all individuals, in particular whether this choice is more liberal or conservative than drawing the sequence of daily time activity patterns at random from the population as a whole. Page 54: The report does: not explain how to interpret the person-day estimates, in relation to setting NAAQS, nor about how the aggregates were computed. In the case of person-days, was this done by: first calculating the expected number of days of exposure per person based on a number of APEX runs; and then multiplying by sub-population size? Some statement of the reliability of these estimates should have been given. Page 79-80: APEX's underestimation of true exposure is of concern as is the use of weekly the aggregation of exposures because the impact on the latter of the ecologic effect (that is not addressed in the report). The errata I pointed out were also ignored in the Jan 07 revision. Page 21: Clock hour "i" rather than "I" is correct, a result of the software's propensity to capitalize "i" whenever it appears alone without quotations around it. Page 40: Section 3.8.2 referred to on this page does not exist even in the final Staff report, so it was not clear what fractions were actually taken from Appendix A and used. MEMORANDUM: "Analysis of uncertainty in ozone population exposure modeling" by John Langstaff Page 8. The common practice of log transforming data from heavy tailed distributions leads naturally to GM =exp (AM) and GSD =exp (SD), and asymmetric 95% confidence intervals for both, (GM/GSD2~ GM*GSDZ) for the GM for example. Therefore, I remain puzzled by the discussion of symmetric intervals on this page, a discussion that I commented about in the Draft. Page 14. The final reportrepeats an earlier error, calling the cross-validation method the "jackknife." CHAPTERS Page 5-18: The analysis in Section 5.3.1.3 is a major improvement over the Drag report. The concern for model uncertaainty has been well addressed and honestly reported. The sensitivity to prior model probabilities'shown in Figure 5-3 at low ozone concentrations correctly reflects the relevant lack of data at that end of the exposure spectrum and points to the need for more experiments in future work to better characterize that part of the exposure response curve. V L~ C-33