Expressed in the mass domain, the LOD90 is the lowest amount of a particular substance for which there is 90% confidence that a single measurement will have a true detection probability of at least 90% while the true non-detection probability of a process blank is at least 90%. The true LOD90 is difficult to measure accurately and must be estimated from the data provided. The data are assumed to be generated from the fully specified ASTM Test Method under a state of statistical control.

The LOD90 value is based on a confidence limit of 0.10. Other confidence limits can be used. Typical values for the confidence limit are 0.10, 0.05, and 0.01 corresponding to LOD90, LOD95 and LOD99, respectively. For convenience, the documentation will simply refer to LOD90.

Upper Confidence Limit for LOD90

At 90% confidence, the possible upper value for the true LOD90 given uncertainties in the assumptions of the predictive model and the limitations in the data provided. This measure of uncertainty does not include uncertainty in the inputted mass values.

Critical Value (CV90)

Expressed in the signal domain, the CV90 is the lowest instrument response for which there is 90% confidence that the response signifies the true detection of a particular analyte. The CV90 value is somewhat below the average response value expected for sample replicates prepared at the LOD90 mass level, and should relate to the value of the peak detection (or alarm) threshold that is set in the trace detector assuming process blank data are realistic.

As with the LOD, different confidence limits can be used. The typical confidence limits of 0.10, 0.05 and 0.01 correspond to CV90, CV95 and CV99, respectively.

90%-90% Upper Tolerance Limit

There are two significance levels given for a tolerance limit. The first level is referred to as the confidence and the second level as the coverage. The 90%-90% upper tolerance limit is interpreted as the value for LOD90 where, at 90% confidence, at least 90% of the possible estimates of the true LOD90 lie below. Further information may be found in the NIST/SEMATECH Engineering Statistics Handbook.

Grubbs Outlier Test

A statistical test used to detect dubious response data (outliers) from a normally distributed population of data. Further information may be found in the NIST/SEMATECH Engineering Statistics Handbook.

Scalable Vector Graphics (SVG) format