I have a random variable $x \in [a, b]$ with PDF $f(x)$ and an event $E$ which satisfies the following property for any $x'<b$.

$$\Pr[E\mid x > x'] \geq \Pr[E]$$

My question is whether or not the following inequality holds.

$$\int_a^b uf(u)\Pr[E\mid x=u] \, du \geq \Pr[E]\int_a^b uf(u) \, du$$


The answer is yes. Indeed, let us write $X$ instead of $x$, according to standard notation, to distinguish between random variables (denoted by upper-case letters) and their values (denoted by lower-case letters). Let us write $P$ instead of $\Pr$, and then let us also write $A$ instead of $E$, to distinguish it from the expectation sign.

Then we need to show that \begin{equation} EX\,P(A|X)\ge P(A)\,EX \end{equation} given that \begin{equation} P(A|X>y)\ge P(A) \end{equation} for all $y$ such that $P(X>y)\ne0$.

Replacing $X$ by $X-a$, we may assume that $X\ge0$. Then \begin{multline*} EX\,P(A|X)=EX\,E(1_A|X)=EX\,1_A=E\Big(\int_0^X dy\Big)\,1_A \\ =E\Big(\int_0^\infty dy\,1_{X>y}\Big)\,1_A =E\Big(\int_0^\infty dy\,1_{X>y}1_A\Big) \\ =E\int_0^\infty dy\,1_{X>y,A}=\int_0^\infty dy\,E1_{X>y,A} \\ =\int_0^\infty dy\,P(X>y,A) =\int_0^\infty dy\,P(A|X>y)P(X>y) \\ \ge\int_0^\infty dy\,P(A)P(X>y)=P(A)\int_0^\infty dy\,P(X>y)=P(A)EX, \end{multline*} as claimed.

(The last equality follows immediately from the special case $EX=\int_0^\infty dy\,P(X>y)$ of the previously proved equality $EX\,P(A|X)=\int_0^\infty dy\,P(X>y,A)$, with $A$ replaced there by an event of probability $1$.)

| cite | improve this answer | |
  • $\begingroup$ Thanks for the answer! Can you please explain the second step in more detail? How do you get $𝐸 \int_{0}^{\infty } dy 1_{X>y,A}$ and what does that mean? $\endgroup$ – Melika May 20 '19 at 18:21
  • $\begingroup$ I have added the details you requested. The meaning is this: $Y:=\int_0^\infty dy\,1_{X>y,A}$ is a random variable (r.v., which is function of the r.v. $X$), and $E\int_0^\infty dy\,1_{X>y,A}$ is the expectation of the r.v. $Y$. $\endgroup$ – Iosif Pinelis May 20 '19 at 20:41

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.