0
$\begingroup$

I stumbled upon a definition of a heavy tail (on the left side) as follows

$\lim_{x\to\infty}( [1-F(x)]e^{\lambda x})=\infty$

for $\lambda>0$. Now on-to questions:

  1. what does it mean if I get $-\infty$? Is it still not heavy-tailed? What if I get zero?

  2. Is this statement equivalent to - only these distributions have infinite moments?

  3. Is this definition of heavy-tail correct for left side tail?

$\lim_{x\to-\infty}( F(x)e^{\lambda x})=\infty$

$\endgroup$
0
$\begingroup$
  1. what does it mean if I get $-\infty$? Is it still not heavy-tailed? What if I get zero?

You cannot get $-\infty$ since cumulative distribution function is a probability, and it cannot increase $1$.

  1. Is this statement equivalent to - only these distributions have infinite moments?

It is equivalent to: only this distributions have infinite exponential moments $\mathbb Ee^{\lambda X}=\infty$ for any $\lambda >0$.

Say, Pareto distribution with probability density function $f(x)=\frac{5}{x^6}$ for $x\geq 1$ has four finite moments $\mathbb EX^4<\infty$ but higher moments are infinite.

The distribution with CDF $F_X(x)=1-e^{-\sqrt{x}}$, $x\geq 0$ has all power moments finite: $\mathbb EX^k<\infty$ for any natural $k$, but $$\lim_{x\to\infty}( [1-F(x)]e^{\lambda x})=\lim_{x\to\infty}e^{-\sqrt{x}}e^{\lambda x}=\infty$$ for any $\lambda >0$.

  1. Is this definition of heavy-tail correct for left side tail? $$\lim_{x\to-\infty}( F(x)e^{\lambda x})=\infty$$

If you mean "for any $\lambda <0$" then it is correct and is equivalent to non-existance of exponential moments $\mathbb E e^{\lambda X}$ for $\lambda<0$. Note that for $\lambda>0$ the condition $$\lim_{x\to-\infty}( F(x)e^{\lambda x})=\infty$$ is impossible: both multipliers tend to zero while $x\to-\infty$.

$\endgroup$

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.