Poisson Prediction Limit
Poisson Prediction Limit
See Also:
Non-Parametric Prediction Limit
Parametric Prediction Limit with Resampling
Description:
The 1992 guidance recommends the Poisson prediction limit for greater than 90% non-detects. The sum of sample concentrations from all background well samples for the specified parameter are used to determine the Poisson prediction limit. The number of recent sampling dates to compare to the prediction limit is specified. Sample concentrations from that number of dates are summed, and the sum compared to the prediction interval. The procedure is repeated for each compliance well.
The 1992 guidance recommends replacing non-detects with ½ the detection limit to implement this method. This can be done from Options | Transform Data.
ChemStat calculates the Poisson prediction limit at both the 99% and 95% levels of significance as both an intra-well and inter-well comparison.
Use:
The method is recommended for situations where there are greater than 90% non-detects. ChemStat compares the prediction limit value to a recent number of recent sampling dates (selected from Select | Recent Dates) for each compliance well. However, the method can also be used to determine a level of triggering for samples that have not yet been collected.
Remarks:
Unlike parametric prediction limits, the method has the disadvantage that if the number of recent dates is greater than 1, it can not be determined which sample caused the statistically elevated result.