2
$\begingroup$

I would like to find the simplest example of a probability density function (pdf) $p(x)$, defined over a support $\sigma$ included in $\mathbb{R}$ (possibly infinite), such that the average $\langle \log(x+2)\rangle$ does not exist (i.e. the integral $\int_\sigma dx\ p(x)\log(x+2)$) does not converge, but $\int_\sigma dx\ p(x)=1$ and $p(x)\geq 0$ for all $x\in\sigma$). Most of the common examples of pdf's (e.g. those for which the Law of Large Numbers does not hold) I can come up with do not seem to do the job. For example, the Cauchy distribution truncated between $1$ and $+\infty$ [$p(x)=(4/\pi)(1+x^2)^{-1}$] is such that $$ \int_1^\infty dx\ p(x)=1 $$ $$ \int_1^\infty dx\ xp(x)=\infty\ , $$ but $$ \int_1^\infty dx\ \log(x+2)p(x)=1.753...\ . $$ Can you help me find a simple example (nothing too fancy, please)? Many thanks for your help, folks.

$\endgroup$

1 Answer 1

3
$\begingroup$

The integral $$\int_0^\infty \frac{1}{(x+2)\ln^2(x+2)}\,dx$$ converges, indeed can be evaluated explicitly by letting $u=\ln(x+2)$. So by scaling by a suitable constant we can produce a density function with support $[0,\infty)$. However, $$\int_0^\infty \frac{\ln(x+2)}{(x+2)\ln^2(x+2)}\,dx$$ diverges.

$\endgroup$
2
  • $\begingroup$ Beautiful (+1). I'll wait a bit more to see if some other examples (hopefully even simpler) crop up. Meanwhile, thanks a lot. $\endgroup$ Feb 21, 2016 at 18:50
  • $\begingroup$ You are welcome. $\endgroup$ Feb 21, 2016 at 18:50

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .