kurtosis, in statistics, a measure of how much of a variable distribution can be found in the tails. The term kurtosis is derived from kurtos (Greek for “convex” or “humpbacked”). A prevalent misconception is that kurtosis measures the “peakedness” of a distribution; however, the contribution of a central peak or range to kurtosis is often small.
Kurtosis is defined as β2 = (E(x)4 / (E(x)2)2) − 3, where E is the expected value of x. The kurtosis of a distribution can be classified as leptokurtic, mesokurtic, or platykurtic. Leptokurtic distributions are variable distributions with wide tails and have positive kurtosis. In contrast, platykurtic distributions have narrow tails and thus have negative kurtosis, whereas mesokurtic distributions (such as the normal distribution) have a kurtosis of zero.