I am currently reviewing the quantile function for discrete random variables but I am a bit confused.

We use the following definition of the quantile function: $$ \tilde{x}_p = \min\{x \in \mathbb{R}\mid F(x) \geq p\} $$ where $$ F(x)=\operatorname{Pr}(X \leq x) $$

Why is it required to have $F(x) \geq p$ if $F(x)$ only gives the probability of $\{X \leq x\}$?
The $\text{greater than}$ relation confuses me.

Because the Distribution function gives me the probability that my discrete random variable $X$ has a value less or equal to some value $x$.

So I would assume that the quantile function has to give a $x$ for a given percentage $p$ which is less or equal than $p$.
But the definition says that the $x$ has to be the smallest value so that $F(x)$ is greater than or equals $p$?

Does it have something to do with the random variable being $discrete$?
If yes could someone elaborate that?

I read different textbooks but even those have different definitions of the quantile.


1 Answer 1


Consider the random variable $X$, which we define to be constant $0$. It has the distribution $$ F(x) = \begin{cases} 0 & \text{if } x\leq 0 \\ 1 & \text{else} \end{cases}$$

Define the quantile function $$F^{-1}(\alpha) = \inf \{ x\in\mathbb{R} \,: F(x) \geq \alpha \}$$

Consider the median of $X$. It is supposed to be $F^{-1}(1/2)$. We know by intuition that it should be $0$, since that is the only value that $X$ will ever take.

Now compare these sets and their infimums \begin{align} S_0 &= \{x\,: F(x) = 1/2 \} = \emptyset &&\inf S_0 = -\infty \\ S_1 &= \{x\,: F(x)\geq 1/2\} = [0,\infty) && \inf S_1 = 0 \end{align}

So if you only allow equality, and not the greater-than-or-equal relation, you get the wrong result.

This fact stems from the discontinouities in the distribution function $F(x)$ and the convention that it is right-continous. If the distribution function was left-continous, we could do some other variant of the quantile function as well.

  • 2
    $\begingroup$ The $F(x)$ that you have defined in your first displayed equation is not a right-continuous function, and for your $F(x)$, it is not true that $S_1 = \{x\colon F(x) \geq \frac 12\}$ is $[0, \infty)$; it is $(0,\infty)$ since your $F(0) $ equals $0$. $\endgroup$ Mar 10 at 2:11

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.