If $X \sim U(a_1, b_1)$ and $Y \sim U(a_2, b_2)$ where $a_1<a_2<b_2<b_1$ (The first one has a higher standard deviation) how do I compute the density of $Z=X+Y$ using the convolution? I think Im getting confused while doing it. Here is what I have:

Denoting $f_x$ and $f_y$ the pdfs, th convolution is

$(f_x * f_y)(x)=\int_{-\infty}^\infty f_x(x-t)f_y(t)dt$

I divide it into different parts

$\int_{-\infty}^{a_1} f_x(x-t)f_y(t)dt+\int_{a_1}^{a_2} f_x(x-t)f_y(t)dt+\int_{a_2}^{b_1} f_x(x-t)f_y(t)dt+\int_{b_1}^{b_2} f_x(x-t)f_y(t)dt+\int_{b_2}^{\infty} f_x(x-t)f_y(t)dt$

Now im having troubles identifying how to compute the intrgals that are not zero, since the values of $f_X(t)$ are not the same as the ones for $f_X(z-t)$

Any help is appreciated, thanks


Correction: Your expression has reversed $b_1$ and $b_2$. The integral of interest is$\int_{a_2}^{b_2}f_X(x-t)f_Y(t)dt.$ To handle $f_X(t-x)$, requires $a_1\lt x-t\lt b_1$ or $x-a_1\gt t\gt x-b_1$. Therefore the lower limit of the integral is $max(a_2,x-b_1)$ while the upper limit is $min(b_2,x-a_1)$. Furthermore there are limits on x determined by the requirement that the upper limit is greater than the lower limit. These are $b_1+b_2\gt x \gt a_2+a_1$.

  • $\begingroup$ The final integral range will depend on x. $a_1+a_2\le x\le a_1+b_2$ range is $(a_2,x-a_1)$. $a_1+b_2\le x\le a_2+b_1$ range is $(a_2,b_2)$. $a_2+b_1\le x\le b_2+b_1$ range is $(x-b_!,b_2)$. $\endgroup$ – herb steinberg May 15 '18 at 16:21

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.