New approaches to the quantification of the robustness as a figure of merit for analytical systems with regard to validation and comparability

Robustness and susceptibility have been introduced to characterize the effects of small variations of the system variables V on the performance S of an (analytical) system in the vicinity of a given working point. To describe the performance over the entire possible range of the parameter variations integral robustness R(int), roughness L(R) and total robustness R(ges) are defined as additional descriptors. Normalization of all scales to the interval 0 to 1 and appropriate definition of R and L lead to dimensionless data. These figures of merit allow to compare the effects of very different variables or of the performance of different procedures. The characterization of the system by R(int), L(R) and R(ges) requires the (approximate) knowledge of the functional dependence S = f(V) of the system on the respective variable. This function may be obtained from models or from (factorial) experimentation. Examples are given for the validation and comparison of ICP-MS, ICP-AES and AAS.