0
$\begingroup$

Let $X$ be a random variable:Choosing a random number with the following Probability density function: $f(x)= \\ 1+x \ ,if \ \ -1<x \le0 \\ 1-x \ ,if \ \ \ 0<x<1 \\0 \ ,else $

Given $X=x$,We will choose a second number $Y$ as following:

if $x\in(-1,0] $,then $Y \mathtt{\sim} U(-1-x,1+x)$.

if $x\in(0,1)$,then $Y\mathtt{\sim} U(x-1,1-x)$.

Find the Probability distribution of $X-Y$.

I wanted to calculate the Cumulative distribution function,but I really have no idea how...

Is this even correct-

$P(X-Y\le t)=E[P(X-Y\le t|X=x)]=\int_{-\infty}^{\infty}P(Y\ge X-t)f_X(x)dx=\int_{-\infty}^{\infty}(1-F_Y(x-t))f_X(x)dx$

And if so,how can I continue?

Or any other ideas?

(Here-$f_X$ is the Probability density function,and $F_X$ is the Cumulative distribution function).

$\endgroup$
1
$\begingroup$

Your approach is a good start. \begin{align*} P(X-Y \le t) &= E[P(X-Y \le t \mid X)]\\ &= \int_{-\infty}^\infty P(Y \ge x-t \mid X=x )f_X(x) \; dx \\ &= \int_{-1}^0 P(Y \ge x-t \mid X=x) f_X(x) \; dx + \int_0^1 P(Y \ge x-t \mid X=x) f_X(x) \; dx \\ &= \underbrace{\int_{-1}^0 \left( \int_{\max\{-1-x,x-t\}}^{1+x} \frac{1}{2+2x} \; du \right) (1+x) \; dx}_* + \underbrace{\int_0^1 \left( \int_{\max\{x-t,x-1\}}^{1-x} \frac{1}{2-2x} \; du \right) (1-x) \; dx}_{**} \end{align*}

We need to analyze the maximum in the $*$ integral. The maximum equals $x-t \iff x>(t-1)/2$. Since $x \in (-1,0)$, we need to incorporate this for $t \in (-1,1)$.

So if $t \in (-1,1)$ then $*$ equals \begin{align*} \int_{-1}^{(t-1)/2} (1+x) \; dx + \int_{(t-1)/2}^{0} \left(\int_{x-t}^{1+x} \frac{1}{2+2x} \; du\right) (1+x) \, dx= \frac{1}{8} (t+1)^2 + \frac{1}{4}(1-t^2) \end{align*} and if $t< -1$, then $*$ equals $0$ and if $t>1$, then $*$ equals 1/2.

Similarly, we look at $(**)$. Note that the maximum in $(**)$ is $x-t$ as long as $t<1$. We also need this maximum $x-t$ to be less than the upper bound $1-x$, which occurs when $x < \frac{t+1}{2}$.

So if $t>1$, we get $1/2$, and if $t<-1$ we get 0, and if $t \in (-1,1)$, we get $$ (**) = \int_0^{(t+1)/2} \left( \int_{x-t}^{1-x} \frac{1}{2-2x} \; du \right) (1-x) \; dx = \frac{1}{8} (t+1)^2 $$

In conclusion, $$P(X-Y \le t) = \begin{cases}\frac{1}{4} (t+1)^2 + \frac{1}{4}(1-t^2)= (t+1)/2, \quad t \in (-1,1) \\ 0 , \quad t \le -1 \\ 1, \quad t \ge 1. \end{cases}$$

Taking the derivative wrt $t$, we get $f_{X-Y}(t) = \frac{1}{2}, \ -1 < t < 1$.

(I'm wondering if there was an easier way to do this . . . exploiting symmetry or something . . . )

$\endgroup$

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.