About estimating the limit of detection of heteroscedastic analytical systems.

The limit of detection is a fundamental figure of merit in chemical analysis. Many different approaches are available for its estimation, but only few allow analysing data characterized by a significant change of precision with concentration. Among these, the ISO 11843-2 approach is certainly the most suitable. However, its implementation may not look easy to operators needing to develop the necessary spreadsheets by commercial software packages. This paper is aimed to verifying if alternative approaches are available for getting acceptably approximate estimates of the limit of detection (LOD). Pragmatically, they were set-up by simply adapting some of the approaches already available in the literature, but compatible only with homoscedastic analytical systems. The results obtained by these alternative approaches when analysing a series of calibrations relevant to the electroanalytical determination of hexavalent chromium in some water samples were compared to those furnished by the ISO approach. The results allowed confirming the critical role of the numerosity of the available data on the consistency of any estimate. Even when using the ISO approach, reliable estimates were only obtained by performing a number of calibrations usually uncommon under standard routine conditions. In such a situation, the differences between the results obtained by all the examined approaches appear less important.