Suppose $X\sim F(n_1,n_2)$, then, what is the distribution of $Y=\log \frac {\exp(X)}{X}$?

I don't even have a clue about this distribution.

Any helpful suggestions and answers are greatly appreciated!

Following Christoph Hanck's suggestion, I got the answer.

Note that $Y=X-\log X \in [1,\inf)$.

The CDF of $Y$ is $$ F(y)=\mathbb{P}(X-\log X \leq y) $$

Since $f(x)=x-\log x$ monotically increases from $(1,\infty)$ and monotically decreases from $(0,1)$.

Then, we derive the range indicated by $X-\log X \leq y$.

For $$ x-\log x=y $$ we can equivalently obtain $$ -xe^{-x}=-e^{-y} $$

Since $-\frac 1 e\leq-e^{-y}<0$, from the wiki page w.r.t. the Lambert $W$ function, we deduce that there are two roots (as expected).

Finally, we have two roots $$ x_1=W_{-1}(-e^{-y})\\x_2=W_{0}(-e^{-y}) $$

Then, \begin{align} F(y)&=\mathbb{P}(W_{-1}(-e^{-y}) \geq X \leq W_{0}(-e^{-y}) )\\ &=\int_{W_{-1}(-e^{-y})}^{W_{0}(-e^{-y})} f_X(x)dx \end{align} and \begin{align} f(y)&=\frac{\partial F(y)}{\partial y}\\ &=f_X\left[ W_{0}(-e^{-y})\right] \frac{\partial W_{0}(-e^{-y})}{\partial y}+f_X\left[ W_{-1}(-e^{-y})\right] \frac{\partial W_{-1}(-e^{-y})}{\partial y} \end{align}

Thanks for the suggestion from @Xi'an.

A more tractable solution can be found from wiki page



Your Answer

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

Browse other questions tagged or ask your own question.