2
$\begingroup$

I found on the book Casella & Berger: "Statistical Inference" the following theorem (2.3.4):

If $X$ is a random variable with finite variance, then for any constants $a$ and $b$ $$\operatorname{Var}(aX+b)=a^2\operatorname{Var}X$$

This makes me wonder the following problem:

Give an example of a random variable $X$ (with infinite variance) and real numbers $a,b$ such that $$\operatorname{Var}(aX+b)\ne a^2\operatorname{Var}X$$

$\endgroup$
  • 1
    $\begingroup$ What happens if $a=0$? Since $\operatorname{var}(X)=\infty$, do you want to interpret $a^2\operatorname{var}(X) = 0^2\cdot\infty$ as $\infty$ or an arbitrary (not necessarily zero) positive finite real number? $\endgroup$ – Dilip Sarwate Oct 4 '14 at 20:45
  • $\begingroup$ I have learned from measure theory that $0\cdot\infty=0$. $\endgroup$ – juniorstatisticstudent Oct 4 '14 at 20:47
  • $\begingroup$ Measure theory does not teach that. The "$\infty$" in measure theory is a special value attained by a measure, subject to strict and limited rules of application, but in this context it would be a way of asserting that an integral diverges, which is not the same thing at all. $\endgroup$ – whuber Oct 5 '14 at 1:41
3
$\begingroup$

Even though $\text{Var}(X)$ is infinite, $aX$ when $a=0$ is not like $0\cdot \infty$; it's perfectly well defined.

As a result, $\text{Var}(aX+b)$ is perfectly well defined -- in that case, it's not indeterminate.

On the other hand $a^2\cdot\text{Var}(X)$ when $a=0$ is of the form $0\cdot\infty$.

$\endgroup$

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.