In the present work, an alternative method for outlier
detection is proposed, based on an approximation of the
experimental distribution with sound theoretical foundations.
Some methods of exploratory data analysis are considered to
model the empirical distribution.