2
$\begingroup$

Suppose we have a continuous random variable $Y$ and a random Bernoulli variable $T$ such that $P(T=1|Y)$ is monotonically increasing in $Y$.

How can we show that $E[Y|T=1]>E[Y|T=0]$?

To me, it makes intuitive sense, but I can't prove it mathematically.

$\endgroup$
0
$\begingroup$

I've managed to answer the question myself. I'm including the solution here for other people that might have the same question.

I want to show that the quantity below is greater than $0$.

$=E[Y \mid T = 1]-E[Y \mid T = 0]$

$=\int^{\infty}_{-\infty}{yP(Y=y \mid T=1)dy} -\int^{\infty}_{-\infty}{yP(Y=y \mid T=0)dy}$

Let $p = P(T=1)$ and $f$ be the PDF of $Y$. Then:

$=\int^{\infty}_{-\infty}{y\left(\frac{P(T=1 \mid Y=y)f(y)}{p}-\frac{(1-P(T=1 \mid Y=y))f(y)}{1-p}\right)dy}$

$=\int^{\infty}_{-\infty}{f(y)y\left(\frac{P(T=1 \mid Y=y) - p}{p(1-p)}\right)dy}$

$=\frac{1}{p(1-p)}(\int^{\infty}_{-\infty}{f(y)y P(T=1 \mid Y=y)dy} - p\int^{\infty}_{-\infty}{f(y)ydy})$

Let $Z=P(T=1 \mid Y)$. Then:

$=\frac{1}{p(1-p)}(E[YZ] - E[Z]E[Y])$

$=\frac{1}{p(1-p)}Cov(Y,Z)$

We know that $Z$ is increasing in $Y$, and the covariance of a random variable and an increasing function of that random variable is always positive (see reference). Therefore:

$=E[Y \mid T = 1]>E[Y \mid T = 0]$

$\endgroup$

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.