Monitoring accuracy and precision — Improvements by introducing robust and resistant statistics

A method is presented to monitor for accuracy and precision of chemical analyses based on the use of a control reference sample (CRS) and blind duplicates of project samples. The major advantage of the method is that it works with real samples instead of international (or laboratory) reference standards. Thus, it will take into account changes in absolute or relative errors over the whole observed concentration range. For each project and each sample type, respectively, and for all elements analysed, it provides realistic estimates of precision and — if there are any determinations at low concentrations — of the practical instead of the rather meaningless theoretical detection limit. By introducing robust and resistant statistics it is possible to drastically reduce the number of samples necessary for the monitoring procedure. As an additional advantage, these statistics are independent of any distribution model and solely reflect the data structure. A program for the whole monitoring procedure, the Laboratory Control Package (LCP), has been written in FORTRAN. It can be implemented on any personal computer with graphic capabilities.