0
$\begingroup$

Probability integral transform states that if a random variable $X$ has a continuous distribution for which the cumulative distribution function (CDF) is $F_X$, then $F_X(X)$ has a standard uniform distribution, that is, $F_X(X)\sim U(0,1).$

My question is about its pdf instead of cdf.

Question: Given a random variable $X$ and its density function $f(x),$ what is the distribution of $f(X)$?

I have a feeling that $f(X)$ does not have a uniform distribution as density is the derivative of CDF. But I do not know what is the derivative of a uniform distribution.

Improve this question
$\endgroup$
  • $\begingroup$ Why do you say 'if any'? Every random variable has a distribution function. There is not much you can say abut the distribution of $f(X)$. It does not have a standard distribution function in general. $\endgroup$ – Kavi Rama Murthy Dec 28 '19 at 12:21
  • $\begingroup$ For example, if $X$ follows a standard normal distribution, what is $E[f(X)]?$ $\endgroup$ – Idonknow Dec 28 '19 at 12:22
  • $\begingroup$ I reckon the "if any" referred to the fact that the distribution may not be universal (as it was in the case with the cdf). Also, minor quibble: not every $X$ has a density to begin with. $\endgroup$ – Clement C. Dec 28 '19 at 12:22
  • $\begingroup$ @ClementC. In this question, we assume that density function exists. $\endgroup$ – Idonknow Dec 28 '19 at 12:23
1
$\begingroup$

It is not universal, and will depend on $X$, so there is no general statement as was the case for the cdf transformation.

For instance:

  • if $X$ is uniform on $[0,1]$, then its pdf $f$ equals $\mathbf{1}_{[0,1]}$ and $f(X) = 1$ is constant a.s.

  • if $X$ is say a standard exponential distribution, then its pdf is $f(x) = e^{-x}\mathbf{1}_{x \geq 0}$ and $f(X)$ is clearly not a constant r.v.: $$\forall y\in[0,1],\qquad \mathbb{P}\{ f(X) \leq y \} = \mathbb{P}\{ X \geq -\log y \} = y$$

Improve this answer
$\endgroup$
  • $\begingroup$ Though your standard exponential distribution has $f(X)$ neatly with a uniform distribution on $[0,1)$. Others will not be as simple $\endgroup$ – Henry Jan 16 '20 at 10:22

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.