Poisson Tolerance Limit
Poisson Tolerance Limit
See Also:
Non-Parametric Tolerance Limit
Description:
The 1992 guidance recommends the Poisson tolerance limit for greater than 90% non-detects. This method determines the upper tolerance limit at both the 95% and 99% levels of significance. The 1992 guidance recommends replacing non-detects with ½ the detection limit. This can be done from Options | Transform Data.
ChemStat provides two different implementations of the Poisson tolerance interval.
1. Comparison to Background Statistic
The upper tolerance limit is determined from the background measurements. If any sample in a compliance well exceeds the tolerance limit, the sample is considered to be impacted.
2. Comparison to Compliance Limit (MCL or Action Limit)
An upper tolerance limit is calculated for each compliance well. If the tolerance limit exceeds the compliance limit, there is statistical evidence of contamination in the well. If the tolerance limit does not exceed the compliance limit, the well does not show contamination.
Use:
The method is recommended in the 1992 guidance for greater than 90% non-detects. Samples from wells indicated as background are used to determine the upper tolerance limit. The upper tolerance limit is then compared to every sample in the each compliance well for the selected parameter.
Alternatively, the compliance limit can be constructed for each compliance well and compared to a compliance limit such as MCL or action limit.
Remarks:
This method has several limitations:
1. The maximum Poisson count of samples is 420.
2. The maximum values for the smallest degree of freedom (from step 5 above) is 400. If a value of 401 is indicated, the Chi-squared values has exceeded the capacity of the algorithm.