SITE INFORMATION AND CORRESPONDENCE

Home Browse Search

David Stavarek, R.G. June 29, 1998 Page 4 assessment should be interpreted as to be limited to the area within facility boundary. Health risks among off-site receptors associated with exposure pathways such as soil dust originating from the contaminated site, may be assessed only to estimate the magnitude of the partial contribution to the overall health risk. 3. Pooling of data. The Workplan proposes the use of the most recent 12-months of on-and off-site concentration data averaged both temporal and spatially. An average value is not sufficient to show that concentrations are increasing or decreasing as a function of time. It is suggested that data analysis of concentration of contaminants in groundwater include or be preceded by a trend analysis over time(a simple tabular or graphical form would be sufficient). If there is no evidence of trend over time, in particular of increasing concentrations, then the pooling of the data seems reasonable. Also, if results from groundwater sampling and analysis are not equally distributed among wells, but rather indicate one or more contaminated groundwater wells,then describe the procedure to be used to develop a representative concentration for the deterministic risk assessment. 4. Replacement value for censored data. The Workplan proposes the use of one-half the minimum detection limit value as a replacement for all censored data(page 16). This approach is not correct. Quantitation detection limits (QDLs) normally vary across analytical runs. Occasionally,values of QDLs are higher than the values of positively detected analytes in other analytical runs. The overall uncertainty of the data set for a particular contaminant,therefore, is affected by the results generated from the runs with the highest detection limit. It is recommended that each censored data be replaced with half of their respective QDLs for the chosen analytical run and method. 5. Definitions. In the equation for the 95%UCL of the mean (page 17) of normally distributed data, the following changes must be made: - in"mean of the traR&feffaed-data" delete the word transformed, - in"s= standard deviation of the#Fin.44med data" delete the word transformed; - it is suggested to use different symbols for the arithmetic mean and the mean of lognormally distributed data. 6. The statement in the last paragraph in page 17, that". . . datasets that are neither normal nor lognormal . . . are non-parametric' is not correct. Revise the concept, as there are many other parametric functions that are non-parametric. 7. Central tendency estimates from non-parametric data. The explanation of the method (pages 17-18) seems irrelevant. No data set in this project will be analyzed with a non-parametric statistical method. Please identify the data set if this is the case, and if so, it is sufficient with referencing the statistical method of analysis.