0
$\begingroup$

I do know that a power law distribution can extend from 0 to $+\infty$, so due to the shape of the distribution, there is no way to define an average value (this might be a characteristic length scale if I understood it correctly?)

But, what if a power law is truncated? still has no characteristic length?

I think we can definitely calculate $E[x] = \int_{a}^{b} xf(x) dx$, because

$f(x) = (1-\alpha)\frac{x^{-\alpha}}{b^{1-\alpha} - a^{1-\alpha}}$.

Is the $E[x]$ a characteristic length?

$\endgroup$

1 Answer 1

0
$\begingroup$

The problem with power law is that integral of the type $$ \int_0^{+\infty}\frac{1}{x^\alpha} dx $$ diverges in infinity, if $\alpha<1$ and diverges at zero, if $\alpha>1$. Thus, the powerl law cannot be normalized, and also has divergent moments. However, it is trancated at the limit where it diverges, it becomes normalizable, although it may still have divergent moments or other averages. If trancated at both limits, all the moments are finite, as can be easily shown by direct calculation.

$\endgroup$
3
  • $\begingroup$ So, if it is normalizable, it has a characteristic length scale? $\endgroup$ Commented Oct 23, 2021 at 0:17
  • $\begingroup$ @TingchangYin yes... but not in the same sense as we put in when saying that a power law doesn't have a characteristic scale. $\endgroup$
    – Roger V.
    Commented Oct 23, 2021 at 4:35
  • $\begingroup$ thanks. I still feel confused about characteristic length. As I understood, e.g. a non-standard normal distribution has a mean $\mu$, and most numbers cluster around $\mu$, so $\mu$ could be a characteristic length of this distribution. A uniform distribution could has a characteristic length which is the average of upper and lower limits. But for truncated power law, is the calculated $E[x]$ a characteristic length , and able to be a typical value representing the distribution? $\endgroup$ Commented Oct 23, 2021 at 9:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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