Let X be a random variable which follows an exponential distribution with parameter $\lambda$ ($\lambda>0$), find the distribution of the random variable $Y = −3\ln(X)$.

So this is my answer for that

$P(Y\leq y) = P(−3lnX \leq y) = P(X\geq e^{-\frac{y}{3}}) = 1-F_x(e^{-\frac{y}{3}})$

$f_Y(y) = \frac{\lambda}{3} e^{-\frac{y}{3}} e^{-\lambda e^{-\frac{y}{3}}}, -\infty \leq y \leq \infty$

We have this definition

If T follows an Gumbel distribution with parameters $\alpha$ and $\beta$, with Probability Density Function $$f_T(t)=\frac{e^{-(t-\alpha)/\beta}}{\beta}e^{-e^{-(t-\alpha)/\beta}}, -\infty \leq t \leq \infty$$

So Y follows Gumbel distribution with parameters $\alpha=0$ e $\beta=3$?

  • $\begingroup$ This is correct if $\lambda =1$. $\endgroup$ – Kavi Rama Murthy Apr 12 at 5:46

Your PDF: $f_Y(y)=\frac{\lambda}{3} e^{-y/3}e^{-\lambda e^{−y/3}}$. All good

Target PDF: $f_T(y) = \frac{e^{-(y-\alpha))/\beta}}{\beta}e^{-e^{-(y-\alpha)/\beta}}$.

So we would like $\lambda e^{-y/3} = e^{-(y-\alpha)/\beta}$.

How do we do that? Taking logs (just for exploratory purposes)

$\ln (\lambda) - y/3 = -y/\beta + \alpha/\beta$

So we will want $\beta = 3$, $\alpha = 3 \ln (\lambda)$.

It also follows that $e^{-(y-\alpha)/\beta}/\beta = \lambda/3 e^{-y/3}$. So it all works!

Hence you can fit a Gumbel to it regardless of the value of $\lambda$ in your exponential distribution.


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.