1
$\begingroup$

Let $X$ and $Y$ be two positive random variables defined over $(\Omega,\mathscr{F},\mathbb{P})$. We know that if they are both log-normal then the random variable $Z$ defined as: $$Z:=\ln\frac{X}{Y}$$ is normally distributed.

Is the converse true? Namely if $Z$ is a normal variable defined as above, can we conclude $X,Y$ are both log-normal? Equivalently, if the ratio $X/Y$ is log-normal, does it imply $X,Y$ are so too?

$\endgroup$
2
  • 5
    $\begingroup$ Definitely not, because you make no assumptions about how $X$ and $Y$ are jointly distributed. So just take $Z$ and independently of it choose $Y$ to have any positive distribution you like and define $X=Ye^Z.$ If you intend that $X$ and $Y$ be independent, then please edit your post to indicate that. $\endgroup$
    – whuber
    Jul 18, 2023 at 13:36
  • 4
    $\begingroup$ You may as well remove the lognormal part, by considering $V=\log(X), W=\log(Y)$ and so ask about $Z=V-W$ and normal distributions $\endgroup$
    – Henry
    Jul 18, 2023 at 13:53

1 Answer 1

2
$\begingroup$

Yes. If $X$ and $Y$ are independent positive variables and $X/Y$ is lognormal, then $\ln(X/Y)=\ln X +(- \ln Y)$ is normal. So by Cramer’s decomposition theorem both $\ln X$ and $-\ln Y$ are normal, and $X$ and $Y$ must be lognormal.

$\endgroup$

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.