3
$\begingroup$

Let $X$ and $Y$ be independent, continuous random variables.

The PDF of a sum $W=X+Y$ is given by the convolution formula $$ f_{X+Y} (w) = \int_{-\infty}^{\infty} f_X(x) f_Y(w-x)\,\,dx$$


In this lecture, it is shown a more general way to use the convolution formula to get the PDF of $Z=2X-Y$.

It goes as follows. If $$Z = (2X)+(-Y)$$ Then $$f_{2X-Y}(z) = \int_{-\infty}^{\infty} f_{2X}(x) f_{-Y}(z-x)\,\,dx$$


What I don't really understand is why the argument of the PDF of $-Y$ should be kept in the same form. Shouldn't it be

$$f_{2X-Y}(z) = \int_{-\infty}^{\infty} f_{2X}(x) f_{-Y}(2x-z)\,\,dx \,\,\,\,\text{?}$$


... since $Z=2X-Y$ and then $Y=2X-Z$ ?



Thanks in advance,
SB

$\endgroup$

1 Answer 1

4
$\begingroup$

$X\neq x$ --> The error in your reasoning is to think that $x$ and $X$ stand for the same object: $x$ is just the variable of integration, we can use any letter. In the first case $x$ is a realization/value of $X$ but in the second case $x$ is a realization/value of $2X$ when you write $f_{2X}(x)$.

$\endgroup$
4
  • 1
    $\begingroup$ ... guess I'll stop writing my answer (+1)! $\endgroup$
    – jbowman
    Commented Jul 10, 2014 at 1:48
  • $\begingroup$ Thanks. I got it for $f_{2X}$, but I still didnt get it for $f_{-Y} (z-x)$. What do you guys recommend? Reviewing convolution derivation? $\endgroup$ Commented Jul 10, 2014 at 1:54
  • $\begingroup$ @SamuelBarbosa: it is just notation. Write $a=2x$. It is now easy to accept that $f_{2X-Y}(z)=\int_{-\infty}^{+\infty} f_{2X}(a)\cdot f_{-Y}(z-a)da$? $\endgroup$ Commented Jul 10, 2014 at 2:19
  • $\begingroup$ Got it from discrete case notation. $$P(Z=z) = \sum P(2X=x).P(-Y=z-x)$$ Now I understand what you were trying to say. Thanks! $\endgroup$ Commented Jul 10, 2014 at 3:05

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.