Selective improvements in multiquality products assisted by rotated principal components

Selective improvements in pharmaceutical multiquality products have been assisted by rotated principal components. Orthogonal and oblique rotations of principal components have been applied to evaluate data from two production process studies. The purpose of the presented methods is to locate the process variables (Xʹs) that can be used to vary some quality parameters (Yʹs), with as minimal influence on other Yʹs as possible, followed by explicit illustrations of such X–Y relations. Two procedures involving varimax rotation are used for this purpose. One, evaluate the loadings from varimax performed on all variables (Xʹs and Yʹs). Two, varimax performed on Yʹs only, followed by linear regression of the varimax Y-scores on Xʹs (VMLR or VPLS) to evaluate the effect of Xʹs on certain Y-groups. The VMLR and VPLS were generally found to be more robust methods compared with varimax only. The information derived from the rotation analysis was used to adjust the relevant X-variables in linear regression models developed to estimate precise Yʹs.