View More View Less
  • 1 Faculty of Natural Sciences and Technology Department of Chemistry Ljubljana (Yugoslavia)
  • 2 E. Kardelj University “Jožef Stefan” Institute 61000 Ljubljana (Yugoslavia)
Restricted access

Abstract  

Large bodies of data on a given material from different sources appear a major advantage of intercomparison experiments and a sound basis for a straight statistical evaluation (as is the usual approach in the assignment of recommended or consensus values). On closer inspection of the results, however, it is usually found that only one or two techniques are represented in statistically significant numbers, most frequently AAS and NAA, and their distribution may not be normal. Correct results are sometimes hidden among laboratory means in the lower or upper part of the frequency distribution plots. Conventional statistics neglecting procedural details and the chemistry behind a procedure may result in incorrect assignment from the reported data, which can be affected by unidentified systematic effects. Generally, in the lowest concentration range contamination predominates as the systematic error determining the accuracy of results. Activation analysis being the least vulnerable to this type of error, it is, if applicable, in principle the most likely to give correct answers. Follow-up work was carried out in this laboratory on the IAEA reference samples of Milk Powder A-11, Muscle Tissue H-4, Fish Flesh MA-A-2 and the Soil-5, clearly demonstrating the potential of activation analysis as verification technique. By determining the same element in the destructive and nondestructive mode, and on the basis of two different isotopes, virtually independent approaches are achieved, often allowing the disclosure of different types of errors. Results are also reported for nickel by a new voltammetric technique enabling its determination at the microgram per kilogram level.