Rationale for using multiple regression analysis with complex interferences.

Abstract

Non-specificities and interferences may become complex when they involve the analyte as well as other interfering substances. These non-specificities and interferences are known as analyte-dependent and multi-interferent interferences. Multiple regression analysis has proven valuable in analysing this type of interference, but the theoretical foundation for using multiple regression analysis to study the basic mechanisms of interference has not been explicitly demonstrated. Graph theory can depict and model the basic mechanisms of interferences and the possible interactions. The relationship between the analyte, the interferents, and the response of the instrument to these entities can be approximated by a polymial of order three, which includes partial derivatives and cross-terms. The partial derivatives relate to the different interactions found with the graph theory model. Further, the partial derivatives can be associated with the coefficients in the multiple regression analysis when the respective values of the three variables (analyte, interferent one, and interferent two) are multiplied by one another. One can decide to retain or discard the coefficient of a variable, based on the statistical significance of the coefficient. The respective interactions in the graphic model can then be assembled and the framework of the interference mechanism established.