Statistical Treatments of Data
The AMC Robust Statistics Toolkit was used to provide a calculation, similar to one-way ANOVA, to estimate the underlying within- and between-laboratory standard deviations. Only data from laboratories returning nine or ten results were included.
The following graphs summarise the performance of laboratories over the course of six rounds of ILC. Blue bars represent the within-laboratory standard deviation and red bars the between-laboratory standard deviation.
Following ILC #3 large standard deviations both within and between laboratories were reported for a sample of horse hair (C1). This was attributed to two different morphologies within the sample, presumably with different isotopic signatures. The same problem was not apparent for nitrogen measured on the same samples (N1). With the exception of the horse hair, the target SD of 0.2 permil was achieved both within and between laboratories.
The other significant exception was 4-nitroacetanilide circulated in ILC#6 (C2), highlighting problems generally associated with the analysis of compounds containing nitro-groups.
Results for nitrogen isotopic analysis generally improved over the course of the first four ILCs. However, the inclusion of more challenging materials such as ammonium nitrate (N2) and sodium nitrate (N3) again highlighted difficulties in analysing nitrogen in its highest oxidation state.
These materials also proved problematic for hydrogen (H1) and oxygen measurements (O1 & O2).
Results for hydrogen isotopic analysis have, generally, been poorer than for other elements. Whilst laboratories were able to achieve typical reproducibility of 0.2 permil, the agreement between laboratories was much poorer.
A particularly problematic material was the sample of ammonium nitrate distributed as part of ILC #5 (H1). Although, the sample of cellulose distributed during the same ILC yielded much more encouraging results, organic materials for the subsequent ILC showed no improvement.
Results for oxygen analysis were, generally similar to hydrogen data, with large between-laboratory SDs. The samples of ammonium nitrate distributed as part of ILC #5 (O1) and sodium nitrate distributed as part of ILC#6 (O2) proved problematic and produced the largest between-laboratory SDs observed.
Very large deviations in the results reported by some laboratories (up to 60 permil) have been attributed to confusion regarding the VBDB and VSMOW reporting scales.