23 laboratories were sent samples for the experiment, and 19 have reported their results.
All the laboratories were given numerical codes that were used in all the cross-testing experiments in the first year of the programme. The participants are able to identify their own results using these codes. For the purposes of this report they have also been assigned letter-codes (because single-character codes are needed for the histograms and Mandel plots).
Samples of three materials were prepared and distributed for the experiment by Partner 3. The materials were a colliery shale (Level 1), an industrial slag (Level 2) and a ground blast-furnace sand used as an additive in cement (Level 3).
A sample of about 30kg of the colliery shale was obtained. This was stepwise crushed and reduced, then milled to produce material that passed a 0.125mm sieve. A rotary divider was used to divide the material into 24 laboratory samples of mass about 0.2kg.
A sample of about 20kg of the industrial slag was obtained and processed in the same way as the colliery shale, again to produce 24 laboratory samples containing about 0.2kg of passing 0.125mm material. However this slag probably came from a steel production process where sulfur is extracted from steel: it contained metallic iron which made it difficult to homogenise and which had to be extracted after crushing and before milling.
This last material was already milled to pass a 0.125mm sieve when received by Partner 3. The rotary sample divider was used to produce 24 laboratory samples of about 0.2kg each.
The participants were asked to take two specimens of approximately 1g from each laboratory sample and to carry out one total sulfur determination on each. Because the original samples were milled to pass 0.125mm before dividing into laboratory samples, the participants were not required to carry out the sample reduction process that would normally be a part of the test method. It is also reasonable to assume that the laboratory samples of each material were practically identical. Hence the measures of repeatability and reproducibility given by the experiment are consistent with the definitions of r and R, and are measures of the precision of the analytical method, but do not include the variability that would normally be introduced by the sample reduction process.
(In other cross-testing experiments in this programme, the symbols r1 and R1 are used to represent the repeatability and reproducibility limits that include a measure of the variability due to sample reduction. To obtain data that would have allowed r1 and R1 to be estimated it would have been necessary to distribute larger laboratory samples - 10kg or more - of uncrushed aggregate, and to have required the participants to divide each laboratory sample into two test portions, and crush and reduce each test portion to produce 1g specimens.)
Where participants did not report the required number of determinations for a level, the missing results are shown using "-.---" in the data tables. When some of a laboratory's determinations are missing, the remaining data have been included in the analysis, but there may not be sufficient data to allow the repeatability of the laboratory to be checked.
The method for the determination of total sulfur does not contain any indication of how the test results should be rounded, apart from a general requirement in the document (covering all its test methods) that the mean of two determinations should be rounded to the nearest 0.01%. However, for the purpose of the cross-testing experiment, it was asked that the test results should be reported to the nearest 0.001%. This was to prevent rounding of the data affecting the assessment of the repeatability and reproducibility of the test method.
Laboratory averages are used to calculate the reproducibility of the test method, and to look for biasses that influence the results of a laboratory at more than one level.
Between-specimen ranges are used to calculate the repeatability of the test method, and to assess the repeatability of tests from individual laboratories.
The averages and ranges are shown in the histograms, and the averages are shown in the Mandel plots.
The averages and ranges are also used to test for stragglers and outliers. Where these have been found, they are indicated throughout using a single question mark (?) to indicate a straggler, and a double question mark (??) to indicate an outlier.
The Mandel plots are used to see if any laboratory suffers from biasses that affect the results at more than one level. Thus, for example, if a laboratory uses a faulty procedure, or a standardised solution that has been prepared incorrectly, then this could cause it to obtain results that are biassed in the same direction in all levels.