A NEW APPROACH FOR OUTLYING RECORDS IN BIOEQUIVALENCE TRIALS

It is well known that outliers can have a significant influence on the conclusion of the bioavail- ability/bioequivalence study. In order to counter the limitations of the actuall inferencial methods, the present communication applies some exploratory data analysis (EDA) techniques, such Andrews curves, Principal Component Analysis (PCA) and Projection Pursuit (PP) for detection discordant records. Al- though computer intensive, the above EDA techniques detect one or more outliers in multidimensional space and does not require knowledge about data distribution.