From the graphs of ranges of replicate determinations, it appears that the variation between replicates for laboratory X is higher than or laboratory Y. Pooled values of these statictics are shown in Table 15, and could be compared using an F-test if it was of interest. Looking at the third graph, it can be seen that there is no obvious pattern or trend in the points. However, whereas the ranges of replicate determinations in Figures 12 and 13 are nearly all less than 0,2, many of the differences between laboratories in Figure 14 are much larger than this. This is aspect requires investigation because it implies that the difference between the laboratories depends on the sample. The average difference between the laboratories may be calculated and is shown in Table 15. It may be used to give an indication of the importance of the difference between the laboratories, on the sample. The average difference between the laboratories, but it may not be used to predict the difference between the laboratories that might be obtained when analysing some subsequent sample. Thus with the transformed data, on average In(Y)-ln(X)=0.443, so Y/X = 1,6 , indicating that laboratory Y obtains results, on average, higher than laboratory X by a factor of 1,6. However with some samples the difference is much larger, and on others laboratory X obtains the higher results.