Between-laboratory variation

The results of interlaboratory study II are presented in Fig. 4.5.1. Five sets of results were obtained for the LAS exercise, and four sets for the NPEO exercise. For LAS, the within-laboratory variability ranged between 2 and 8% (RSD) for sample III (distilled water spiked with lmgL-1 LAS), 1 and 13% for sample 112 (wastewater influent), and 3 and 8% for sample 113 (sample 112 spiked with lmgL-1 LAS). Between-laboratory variations (calculated from the mean of laboratory means, MOLM) amounted to RSDs of 15, 30 and 30% for samples III, 112 and 113, respectively. The LAS values reported were in the range of 700—1100 p,g L-1 in sample III, 1100-1800 p,g L-1 in sample 112 and 1900-3000 p,g L-1 in sample 113, indicating that even in the matrix wastewater influent, the spiked concentration of lmgL-1 LAS could be almost quantitatively determined by all laboratories. 

There does not appear to be evidence to demonstrate the variability of either abrasion standards or abradants but within one laboratory the coefficient of variation of abrasion results using different batches of a standard rubber would probably be not much less than 8%, and the between laboratory variation could clearly be very much greater. Some abradants will certainly be more variable than this but other materials can be reproduced with better precision. Although it is a fact that standard rubbers are themselves variable, they are of very considerable value, particularly when reference is made only to standards from one batch and where they are used to monitor the change with time of one sample of abradant or to compare a number of abradants. 

Between-laboratory variation Will the precision and accuracy of the method be the same between the development and quality control laboratories ... 

A common example where ANOVA can be applied is in interlaboratory trials or method comparison. For example, one may wish to compare the results from four laboratories, or perhaps to evaluate three different methods performed in the same laboratory. With inter-laboratory data, there is clearly variation between the laboratories (between sample/treatment means) and within the laboratory samples (treatment means). ANOVA is used in practice to separate the between-laboratories variation (the treatment variation) from the random within-sample variation. Using ANOVA in this way is known as one-way (or one factor) ANOVA. 

Figure 2.2 Different laboratory performances for 30-min Microtox tests with 3,4-dichloroaniline (/) acceptable precision (—) marginal acceptability (x) excessive variability between repeat tests. Values falling between upper and lower 95% confidence intervals indicate acceptable control over between-laboratory variation.

Current inter- and intra-laboratory evaluations indicate the Equilibrium Jar s precision is 8% and between laboratory variation is about 10%. 

Between-laboratory variation was much larger than within-laboratory variation, but there was no apparent bias between the laboratories (Table II). About 20 percent of the between-laboratory differences... 

One of the early pioneers in the analysis of ITP data, W. J. Youden, demonstrated more than thirty years ago the dominant influence of interlaboratory bias (or systematic error as he called it) in a series of publications [4,5,15]. He showed with simple graphical plotting techniques the unmistakable existence of bias. The existence of an essentially constant bias for any laboratory invalidates the customary assumption in ITP analysis that a random normal distribution adequately represents the between-laboratory variation. See Annex D. 

ASTM D 3937 provides a method for determining the crimp frequency of synthetic staple fibers, but the between-laboratory variation is known to be great. It is, however, suitable for in-house QC operations. 

This is a two-way ANOVA experiment with repiication. The mean squares for between-row, between-column, interaction, and residual variations are respectively 2.53 (2 d.f.), 0.0939 (2 d.f.), 0.0256 (4 d.f.), and 0.0406 (9 d.f.). The interaction mean square is less than the residual mean square, so sample-laboratory interactions are not significant. Comparing the between-column (i.e. between-laboratory) and the residual mean squares gives F= 0.0939/0.0406 = 2.31. The critical value of 2,9 is 4.256 (P = 0.05), so the between-laboratory variation is not significant. 

Two types of method validation can be distinguished. Full method validation, of interest to the general scientific community, is carried out through an interlaboratory method performance study. Where a method becomes more routinely used it is reasonable to expect that the method should be fully validated. Internal method validation (single-laboratory method validation) is a scientific and technical alternative. It consists of validation steps carried out within one laboratory, for instance, to validate a new method that has been developed in-house or to verify that a method adopted from some other source is applied sufficiently well. A single-laboratory validation cannot assess between-laboratory variation and will provide an optimistic assessment of interlaboratory variability (cfr. Chp. 6.2.3 and 6.2.4). In-house method validation is described in the lUPAC, AOAC International, and ISO guidance [65,66]. There are several types of internal laboratory validation ... 

---