A comparison of three procedures for robust PCA of experimental results of the homogeneity test of a new sodium diclofenac candidate certified reference material

This paper compares classic and robust multivariate methods for evaluation of experimental results from the homogeneity test of a new sodium diclofenac candidate certified reference material (CRM). The results showed that the robust principal component analysis (PCA) based on projection pursuit was the most effective method for identification of outliers compared to the classic method and to the two other robust approaches: the ROBCA algorithm and the spherical PCA, when the concentrations of all active pharmaceutical ingredient (API) impurities were considered simultaneously. The PCA based on projection pursuit was able to identify six outliers, while the other methods identified only five. Through the use of these statistical tools, it was possible to reduce the value of the standard uncertainty due to between-bottle (in)homogeneity (ubb) and to guarantee an accurate result of the combined standard uncertainty of the certified reference material (uCRM).