We are having a disagreement in reference to the Pooled level of Quantification, limit of quantification, Limit of Detection and how to perform the calibration curves with statistical validity. As the concentration of a substance is reduced the pulses (attribute data) reduce. The pulses at very low concentration theoretically will go to zero and at high concentrations will be say 2000. For example 0 results in 14 with sigma 5, 1 results in 110 with sigma 12, ands 20 results in 2450 with a sigma of 49. (many more data point but this is a trend). So now we are looking at the poisson distribution of the data for calibration curves with the standard deviation of the curve decreasing as a percent of the signal at each data point. This indicates that the uncertainity of the fit curve fit is not constant across the calibration curve. This would infer that the use of the LOD being 3 sigma above the blank and LOQ being 10 sigma above the blank would rely on non-constant sigma’s. So what is the proper LOQ, LOD determination for this situation?