TNI Logo The NELAC Institute

Standard Interpretation



SIR #: 214
Standard: 2009 TNI
Section: V1M6, Section 1.5.4


REQUEST:

We are not sure exactly what this section is requiring of us. What does it mean by "the experimentally observed precision at each testing level"? We are assuming that the simplified version of this section would say that our calculated precision values from our duplicates cannot be greater than the uncertainty of either sample used in the calculation. Is that correct?

Thank you!

TNI FINAL RESPONSE:

(Quality Systems Expert Committee and NELAP AC, 10-7-13)

Section 1.5.4 states specifically that "the experimentally observed precision at each testing level [of the precision evaluation in section 1.5.3] shall not be statistically greater than the maximum combined standard uncertainty of the measurement results at that level, although it may be somewhat less." Section 1.5.3 establishes different approaches for "reference methods" and for laboratory-developed (or modified) methods.

For "reference methods", 1.5.3 a) the standard deviation of the results is calculated for at least four spiked samples as described in Section 1.6. The standard deviation of the four replicate results is compared to the calculated combined standard uncertainty for each of the four results. The CSU is acceptable as long as it is equal to or greater than the experimental standard deviation of the four results.

1.5.3 b) addresses non-grandfathered, non-reference methods. It states that the laboratory shall use a documented procedure to evaluate precision and bias. An acceptable approach will determine the standard deviation for at least three blank samples, and at least three replicate samples at each of three known activities that span the range of activities expected from samples to be analyzed using the method. The calculated CSU for each sample result is compared to the standard deviation of results at that activity level and is considered acceptable as long as it is statistically equivalent, or slightly lower.

Analysis of duplicate samples will not meet the minimum requirements of either of these two approaches.