A heavy-tailed distribution is often defined as a distribution with a tail that is not exponentially bounded.

A truncated power law (or power law with exponential cut-off) is a distribution that consist of a power law distribution multiplied by an exponential distribution:

$$ f(x) \propto x^{\alpha}e^{\beta x} $$

This distribution behaves alike a power law for small values of x, but experiments a heavier decrease in the probability of large values of x than a power law. Some papers describes truncated power laws behaviors as follows:

[The truncated power law] has the power law’s scaling behavior over some range but is truncated by an exponentially bounded tail.

Is the truncated power law a heavy-tailed distribution due to its power law-like behavior over the smaller x range?

Alternatively, is the exponentially bounded tail of the truncated power law enough to reject it is a heavy-tailed distribution?



A main reason that heavy-tailed distributions are of interest is that they can model processes that produce occasional extreme values. Infinite tails are not necessary for this purpose; many (considering our measurement systems that give us actual data) real processes are both bounded and outlier-prone. Truncated distributions with truncation points sufficiently far out in tails satisfy a definition of heavy-tailedness that matches the actual intent of their use.

Classic definitions of heavy-tailedness that require infinite tails are too narrow: They do not admit a large class of very relevant distributions that are very useful for modelling data-generating processes that produce occasional extreme values; namely, bounded distributions.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.