Overcoming constraints of the control charting approach for measurement uncertainty assessments

Abstract

This report proposes an alternative approach for measurement uncertainty estimation avoiding the two main limitations of the control charting approach in a context were certified reference materials are unavailable. The procedures under study estimated the mass fraction of total alkaloids as nicotine and Kjeldahl total nitrogen in powdered samples of air-cured leaves of Cuban cigar tobacco. The intermediate precision estimated from results of different analysts was evaluated in parallel at three levels of mass fraction trough a spiking experiment using test portions of a one composite non-certified reference material per procedure and a balanced fully nested design. The systematic effect component was estimated in the same mass fraction levels but using a different set of samples covering a wide range of the typical routine test materials. An iterative re-weighted least squares linear regression analysis was performed between the results obtained through the assessed procedure with those attained from a reference one which analyses the same measurand. The estimation was then carried out using a proposed mathematic expression derived from the application of the uncertainty propagation law to the equation for bias estimation for every reference concentration. The uncertainty associated with bias estimation was furthermore evaluated over the whole mass fraction range considered in the procedure comparison study. The interpretation of the evaluated parameters is discussed regarding the specifics analytical characteristics of each procedure.