Novel multivariate statistical method for the detection of batches with more variability than others of the same drug

The thesis ended up providing a set of tools for assessing the variability differences between production batches of the same drug in pharmaceutical datasets. Its main feature is the Mahalanobis-Dispersion model which, based in the multivariate distance of Mahalanobis, computes various measures of data dispersion within each batch then, these measures can be statistically tested at a desired significance level. The chosen statistical test is the one-way analysis of variance (ANOVA), it can be performed as usual as well as using a iterative procedure which pops out of the dataset the batch with higher dispersion in each iteration, like what one would do with the Cochran's C test in univariate cases.