5
$\begingroup$

If $X$ is a continuous random variable with support $A$, does this imply that the cdf of $X$ is strictly increasing on $A$?

My guess is yes. But just in case, let me know if you can think of any counterexamples.

$\endgroup$

2 Answers 2

3
$\begingroup$

A counter example is made by taking the cdf as the Cantor distribution $\mathfrak c(\cdot)$ on $[0,1]$, whose support is the Cantor set $\mathfrak C$:

  1. it is a continuous (if not absolutely continuous) distribution
  2. its support $\mathfrak C$ is closed and of Lebesgue measure zero
  3. the points 1⁄3 and 2⁄3 are adjacent in $\mathfrak C$, i.e., $\mathfrak C\cap(1/3,2/3)=\emptyset$ while $1/3,2/3\in\mathfrak C$
  4. hence $\mathfrak c(1/3)=\mathfrak c(2/3)$ and $\mathfrak c(\cdot)$ is not strictly increasing on $\mathfrak C$
$\endgroup$
4
  • 2
    $\begingroup$ Brilliant counter-example. What happens if the density exists as discussed in the other answer? $\endgroup$ Dec 21, 2020 at 10:57
  • 3
    $\begingroup$ @Cross Wherever the density might be zero, the cdf is not strictly increasing. It is possible for densities to equal zero on the support of the random variable--they just cannot be zero throughout any open interval. $\endgroup$
    – whuber
    Dec 21, 2020 at 12:35
  • $\begingroup$ Thanks, Xi'an! This is very helpful. $\endgroup$ Dec 21, 2020 at 22:57
  • $\begingroup$ @Xi'an@whuber Then requiring the density to be everywhere positive would suffice, right? $\endgroup$ Dec 21, 2020 at 22:58
2
$\begingroup$

Assuming the density exists, and ignoring the boundary points, a function strictly increases if its derivative is strictly positive. CDF’s derivative is PDF and it is strictly positive in the support. So, CDF must strictly increase in the support.

$\endgroup$
2
  • 5
    $\begingroup$ A continuous random variable being a random variable whose cumulative distribution function (cdf) is continuous everywhere, this cdf may be differentiable nowhere, meaning there is no density (pdf). $\endgroup$
    – Xi'an
    Dec 21, 2020 at 8:01
  • 1
    $\begingroup$ @gunes Thank you very much! $\endgroup$ Dec 23, 2020 at 20:23

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.

Not the answer you're looking for? Browse other questions tagged or ask your own question.