Given X and Y as two independent random variables following $U(0,1)$ and We are required to obtain the distribution of $Z = X + Y$. The answer is given as follows:

$$f(z) = \begin{cases} z & \text{for $0 < z < 1$} \\ 2-z & \text{for $1 \le z < 2$} \\ 0 & \text{otherwise.} \end{cases}$$

I generally get confused in these types of problems when there is a split in the support of the distribution of interest. Here, I know one thing for sure that $Z$ will have values ranging from 0 to 2. I am not able to understand how do we think through the splitting of these z values into 0 to 1 and then 1 to 2.

Let me show you how would I have approached this problem.

Let $$U = X + Y \text{ and } V = Y,$$

then I can write $$X = U - V \text{ and } Y = V.$$

I am trying to find the joint distribution of U and V by jacobian transformation. The value of jacobian will be 1.

$|J| = 1$

Now, we can write:

$f(u,v) = g(x,y)|J| = g(x,y) = g(u-v,v) = 1,\text{ for } 0<u-v<1, 0<v<1$

Now, to get the marginal of $U$, I would need to integrate this joint with respect to $v$. We can see that v takes value from 0 to 1 but at the same time $v>u-1$. So, we can integrate the above joint density with respect to $v$ from $u-1$ to 1 and we get $f(u) = 2-u$.

Now, I have a question here that where exactly in the above steps, I would have thought of splitting that support of z and why? Any help would be appreciated.

  • $\begingroup$ I was thinking about this. One thing that I came up with is: since $0<u-v<1$, it implies that $v<u<1+v$. Also, $v$ itself is between 0 and 1. I can write $0<v<u<1+v<2$. Hence, from this new modified limits, I can see two different ranges of values that v can take. One is $0<v<u$ and another is $u-1<v<1$. Is this reasoning correct for this problem? $\endgroup$
    – userNoOne
    Feb 12 at 6:18
  • $\begingroup$ Please add self-study as a tag. $\endgroup$
    – Xi'an
    Feb 12 at 7:20
  • 2
    $\begingroup$ For an account of several ways to deal with this issue, see stats.stackexchange.com/a/43075/919. $\endgroup$
    – whuber
    Feb 12 at 15:16

The joint density is $$f(u,v)=\mathbb I_{(0,1)}(u-v)\mathbb I_{(0,1)}(v)$$ For a fixed value of $u$, the function $$v\longmapsto \mathbb I_{(0,1)}(u-v)\mathbb I_{(0,1)}(v)$$ can be rewritten $$v\longmapsto \mathbb I_{0\le u-v\le 1}\mathbb I_{(0,1)}(v) =\mathbb I_{u-1\le v\le u}\mathbb I_{(0,1)}(v)=\mathbb I_{(u-1,u)}(v)\mathbb I_{(0,1)}(v)$$ Meaning that, conditional on $U=u$, $V$ is between $0$ ans $1$ AND between $u-1$ and $u$. Thus$$ \mathbb I_{(0,1)}(u-v)\mathbb I_{(0,1)}(v)=\mathbb I_{(\max(0,u-1),\min(u,1))}(v)$$

  • $\begingroup$ What does the symbol $\longmapsto$ mean here? $\endgroup$
    – userNoOne
    Feb 12 at 8:06
  • $\begingroup$ This is a traditional mathematical notation for functions and reads "maps $v$ to $\mathbb I$..." $\endgroup$
    – Xi'an
    Feb 12 at 8:45

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.