We need to establish the given inequality : $$F_X(x) + F_Y(y) - 1 \leq F_{X,Y}(x,y) \leq \sqrt{F_X(x) F_Y(y)}$$ where $X$ and $Y$ are any random variables.

First I tried the R.H.S :

I started working on p.d.f's ( probability density functions first ).

$ f_{X|Y}(x|y) = \dfrac{f_{X,Y}(x,y)}{f_Y(y)}\implies \dfrac{f_{X|Y}(x|y)}{f_X(x)} = \dfrac{f_{X,Y}(x,y)}{f_Y(y)f_X(x)}$

But I don't think that's gonna work . Can anyone suggest some other way ?

  • $\begingroup$ Since $X$ and $Y$ are said to be any random variables, using density functions will at nest lead to a proof restricted to jointly continuous $X$ and $Y$. @StubbornAtom's answer is very much to the point (+1 to him) and deserves acceptance. $\endgroup$ Mar 19, 2020 at 14:57

2 Answers 2


Hint for the right hand side:

$F_X(x) = P(X < x) \ge P(X < x \wedge Y < y) = F_{X, Y}(x, y)$.

Hint for the left hand side:

The following figure shows the $XY$ plane, with the horizontal and vertical lines intersecting at $(x, y)$.

enter image description here

The region $C$ is the area where $X < x \wedge Y < y$ - you need to integrate over it to obtain $F_{X, Y}(x, y)$. Over which regions do you need to integrate to obtain $F_X(x)$, $F_Y(y)$, and $1$?

  • $\begingroup$ How did you conclude the first one , i.e , $P(X<x) \geq P(X<x , Y<y)$ ? $\endgroup$
    – User9523
    Oct 16, 2016 at 18:44
  • 1
    $\begingroup$ $P(X < x \wedge Y < y) = P(X < x) P(Y < y | X < x)$. A conditional probability is nevertheless a probability. What is the range of a probability? $\endgroup$
    – Ami Tavory
    Oct 16, 2016 at 18:45
  • $\begingroup$ You're welcome. All the best. $\endgroup$
    – Ami Tavory
    Oct 16, 2016 at 18:53

It is just the elementary inequality $$P(A)+P(B)-1\le P(A\cap B)\le \sqrt{P(A)P(B)}$$ for events $A=\{X\le x\}$ and $B=\{Y\le y\}$. There is no need to go into distributions.

Let $I_A$ be the indicator of $A$, i.e. $I_A=1$ if $A$ occurs and $I_A=0$ if $A$ does not occur.

Then by Cauchy-Schwarz inequality,

$$\left(E\left[I_AI_B\right]\right)^2 \le E\left[I_A^2\right]E\left[I_B^2\right] \implies (P(A\cap B))^2\le P(A)P(B)$$

And $P(A^c\cup B^c)\le P(A^c)+P(B^c)=1-P(A)+1-P(B)$ implies $$P(A\cap B)=1-P(A^c\cup B^c)\ge P(A)+P(B)-1$$


Your Answer

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

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