We have X and Y, independent, zero mean random variables. Show that:

$$E|X+Y| \geq E|X|$$

What I tried:

I treated them as continuous variables, though it should be true in general. But I could not derive the proof. I found that it should be proven:

$$\int_{-\infty}^{\infty}\int_{-x}^{\infty}(x+y)f(x)g(y)dydx \geq \int_0^{\infty} xf(x)dx$$

where we have the constraints:

$$\int f(x)dx=1$$

$$\int x\cdot f(x)dx = 0$$

$$\int g(y)dy=1$$

$$\int y\cdot g(y)dy = 0$$

  • 1
    $\begingroup$ Do you know Jensen's inequality? $\endgroup$ – kimchi lover Jul 20 '17 at 12:10
  • 2
    $\begingroup$ This was asked before. Use the fact that, by convexity, $$E(|x+Y|)\geqslant|x+E(Y)|=|x|$$ for every $x$, and integrate this inequality with respect to the distribution of $X$. $\endgroup$ – Did Jul 21 '17 at 12:58
  • $\begingroup$ @kimchilover I see what you are hinting at but, to show $E(|Z|)\geqslant|E(Z)|$, Jensen is hardly necessary. $\endgroup$ – Did Jul 21 '17 at 18:57
  • 1
    $\begingroup$ @Did could you show me please, where it was asked? Thank you! $\endgroup$ – AvanDavad Jul 27 '17 at 14:28
  • 1
    $\begingroup$ @Did you are totally right, I did the search before posting but I did not find it... $\endgroup$ – AvanDavad Jul 28 '17 at 16:37

We have that

$$|EX| \leq E|X|$$

which is true in general.

$$E(abs(X+Y)| X) \geq |E(X+Y|X)| = |X + E(Y|X)| = |X|$$

take the expected value (and use the tower rule, aka. the law of total expectation):

$$EE(abs(X+Y)|X)\equiv E|X+Y| \geq E|X|$$


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.