Estimation of measurement uncertainty in chemical analysis
10.2. Uncertainty component accounting for random effects
Estimating the within-lab reproducibility component of uncertainty
http://www.uttv.ee/naita?id=17913
https://www.youtube.com/watch?v=9-803-VViKo
The within-lab reproducibility (intermediate precision, sRW, denoted as u(Rw) in the Nordtest guide) component takes into account all uncertainty sources that are random in the long term (i.e. several months, preferably one year). So, quite some uncertainty sources that are systematic within a day will become random in the long term.
This is one of the key points of the within-lab validation approach of uncertainty estimation: trying to account for as many as possible uncertainty sources via random effects. This is also an important advantage of the approach, because as a rule estimating random effects can be done more reliably than estimating systematic effects. The reason is that for evaluating random effects there is no need for a reference value, but for evaluating systematic effects there is (see section 10.3).
The simplest way to find u (Rw) is from a number of repeated measurements of a control sample, organized, e.g. as a control chart. Alternatively, the pooled standard deviation approach as explained in section 6 can be used. If this is done then the u (Rw) can be found based on several different control samples, so that it will be an average value of all of them.
The number of values used for evaluation of u (Rw) must be sufficiently large. An initial estimate of u (Rw) can be obtained with 10-15 values but thereafter more data should be collected. Even more importantly, the time period during which the data are collected, must be sufficiently long (at least several months, preferably around a year) so that all the sources of variability in the procedure are taken into account. So, 10 values collected over a five-month time period is a better option than 20 values collected during 1.5 months. If the measurements are done with the same control sample, then it must be available in sufficient amount and must be stable during the time period. Table 6.2 in section 6 gives a compact overview of the requirements of u(Rw) (sRW) determination.
Depending on situation u (Rw) can be used as absolute or relative value.
It is important to stress that u (Rw) should be estimated separately for different matrixes and different concentration levels.
