Analytical results without systematic errors and with accurately known random errors are normally distributed around their true values. Such results may be produced by means of Neutron Activation Analysis both with and without radiochemical separation. When all sources of random variation are known a priori, their effect may be combined with the Poisson statistics characteristic of the counting process, a.id the standard deviation of a single analytical result may be estimated. The various steps of a complete neutron activation analytical procedure are therefore studied in detail with respect to determining their contribution to the overall variability of the final result. Verification of the estimated standard deviation is carried out by demonstrating the absence of significant unknown random errors through analyzing, in replicate, samples covering the range of concentrations and matrices anticipated in actual use. Agreement between the estimated and the observed variability of replicate results is then tested by a simple statistic T based on the chi-square distribution. It is found that results from neutron activation analysis of biological samples can be brought into statistical controlIn routine application of methods in statistical control the same statistical test may be used for quality control when some of the actual samples are analyzed in duplicate. This Analysis of Precision serves to detect unknown or unexpected sources of variation of the analytical results, and both random and systematic errors have been discovered in practical trace element investigations in different areas of research. Particularly, at the ultratrace level of concentration where there are few or no Standard Reference Materials for ascertaining the accuracy of results, the proposed quality control based on the Analysis of Precision combined with neutron activation analysis with radiochemical separation, with an a priori precision independent of the level of concentration, becomes a powerful tool for controlling accuracy too.
|Place of Publication||Roskilde|
|Publisher||Risø National Laboratory|
|Number of pages||300|
|Publication status||Published - 1980|
|Series||Denmark. Forskningscenter Risoe. Risoe-R|