Generally speaking, inferential statistics focuses on the analysis of a sample taken from a population in order to draw mathematically founded conclusions about the population. In the majority of the applications, our interest is focused on the central part of the population, i.e., the first, or perhaps the first two moments. This is true both for traditionally presented simple indicators, and more complicated models (e.g., regression, which focuses on the first – conditional – moment) too. The significance of skewed distributions or outlying observations is recognized, but the usual approach is to somehow eliminate them (e.g., by transforming the distributions, using other, robust statistics, or removing outliers). Rarely is the opposite attempted, i.e., getting rid or putting less weight on the values from the center and investigating extreme observations, despite the fact that ”outliers” – when they’re not results of data entry errors or sensor malfunction – could contain very valuable information.