Defining limits for precision

Residual maximum likelihood analysis of the ring-test data also provides an estimate of the underlying repeatability of each test method ( repeatability)- As part of a quality control scheme, this can be compared with variances estimated from the EC/LC50 values derived from a series of repeat tests carried out within a laboratory using the same toxicant (S2) and using a conventional 2 test, as shown below  

If S2 is significantly greater than o-2epeatabili)y for that test method, excessive variability between repeat tests is indicated and an investigation of possible sources of error might be prompted to restore within-laboratory variability to an acceptable level. 

In this example, each laboratory has performed six tests. Laboratories A, C, D, F and G would meet the criteria for both within- and between-laboratory variability. By contrast, data from laboratory J would be acceptable in terms of the between-laboratory criterion but the poor agreement between EC50 values would marginally fail the criterion for within-laboratory variability. In this example, data from laboratories H and I exhibit poor repeatability and also fail the criterion for between-laboratory variability. 

9 Test method development and the derivation of quality control 

The development of new test guidelines or the revision of existing ones provides an opportunity to address quality control considerations as well as the more conventional considerations concerned with the technical conduct of the tests. Although ring-testing is frequently involved, rarely is it designed in such a way 

---