Let's say we have two independent random variables, $X$ is uniform on $[0,1/2]$ and $Y$ is uniform on $[1/2,1]$. If we look at the distribution of $X+Y$, is it triangular distribution between $[1/2,3/2]$ where the peak is $1$?

I used to same logic as in the calculation of two identical uniform random variables but my calculations give a different result. If p.d.f. of $X$ and $Y$ is $f$ and $g$, respectively, \begin{align} f_{X+Y}(z)=\int_{-\infty}^{\infty} f(z-y)g(y)dy \end{align}

If $y\leq 1/2$ or $y\geq 1$, $g(y)=0$. So, \begin{align} f_{X+Y}(z)&=\int_{1/2}^{1} f(z-y)2 dy\\ \end{align}

If $0\leq z \leq 1/2$ or $z\geq 3/2$, $f(z-y)=0$.

If $1/2\leq z \leq 1$, $f(z-y)=1$ if $1/2 \leq y \leq z$. Then, \begin{align} f_{X+Y}(z)&=\int_{1/2}^{1} f(z-y)2 dy\\ &=\int_{1/2}^{z} 2*2 dy\\ &=4z-2 \end{align}

If $1/2\leq z \leq 3/2$, $f(z-y)=1$ if $z-1 \leq y\leq 1$. \begin{align} f_{X+Y}(z)&=\int_{z-1}^{1} f(z-y)2 dy\\ &=\int_{z-1}^{1} 2*2 dy\\ &=8-4z \end{align}

Could you tell me if my calculations are correct? If they are, I still feel like the result should be triangular ad I cannot see why not.


The first integral is correct; the second is incorrect. You should write that if $1 < z \le \tfrac{3}{2}$, then $f(z-y) = 2$ if $z - \tfrac{1}{2} \le y \le 1$. This gives the correct density $f_{X+Y}(z) = 6-4z$ over $1 < z \le \tfrac{3}{2}$.

A simpler way to do this calculation is to observe that $X^* = Y - \tfrac{1}{2} \sim {\rm Uniform}(0,\tfrac{1}{2})$, so the distribution of $X+Y$ is equal to the distribution of $X + X^* + \frac{1}{2}$, where $X$ and $X^*$ are IID.


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.