Transforming random variables and finding new bivariate and marginal distribution

88 Views Asked by At

Consider the following bivariate distribution

$$ f(x,y) = \begin{cases} \frac{3}{28}(x^2 + xy) \quad &\text{for } 0 \leq x \leq 2 \text{ and } 0 \leq y \leq 2 \\ 0 &\text{otherwise} \end{cases} $$

and the following two new random variables

$$ U = X + Y, \quad V = X $$

Find the new bivariate distribution of $U$ and $V$, and find the marginal distribution of $U = X + Y$.

I'm having some trouble figuring out the second half of the question, and I would appreciate any help I could get. I think I've managed to find the new bivariate distribution. Here's what I've done so far:

First I express the original random variables in terms of the new ones

$$ x = v_1(u,v) = v, \quad y = v_2(u,v) = u - v $$

which can be inserted into the following formula to find the new bivariate distribution:

$$ g(u,v) = f(v_1(u,v),v_2(u,v)) \cdot |\det(M)| $$

$$ \det(M) = \frac{\partial v}{\partial u} \cdot \frac{\partial (u - v)}{\partial v} - \frac{\partial v}{\partial v} \cdot \frac{\partial (u - v)}{\partial u} = -1 $$

Before putting all of this together I also need to find the new region of positive density. What I found was

$$ 0 \leq x \leq 2 \implies 0 \leq v \leq 2 \\ 0 \leq y \leq 2 \implies 0 \leq u - v \leq 2 \implies v \leq u \leq v + 2 $$

and putting all of this together yields:

$$ g(u,v) = \begin{cases} \frac{3}{28} uv \quad &\text{for } v \leq u \leq v + 2 \text{ and } 0 \leq v \leq 2 \\ 0 &\text{otherwise} \end{cases} $$

As far as I understand, to find the marginal distribution of $U$ I need to evaluate

$$ f_U(u) = \int_{-\infty}^{\infty} f(u,v) dv $$

over the following region of positive density. This this looks very different to the other examples I've seen in my book, and I'm not sure how to proceed here.

1

There are 1 best solutions below

4
On BEST ANSWER

Let's begin with

$$f_U(u)=\int_0^u \frac{3}{28}uv dv=\frac{3}{56}u^3\mathbb{1}_{[0;2)}(u)$$

$$f_U(u)=\int_{u-2}^2 \frac{3}{28}uv dv=\frac{3(4-u)u^2}{56}\mathbb{1}_{[2;4]}(u)$$

So the marginal U has the following density

$$f_U(u)=\frac{3}{56}u^3\mathbb{1}_{[0;2)}(u)+\frac{3(4-u)u^2} {56}\mathbb{1}_{[2;4]}(u)$$

The marginal V is easier

$$f_V(v)=\int_v^{v+2} \frac{3}{28}uv du=\frac{3v(v+1)}{56}\mathbb{1}_{[0;2]}(v)$$

This because you have

$0\leq u-v \leq 2$

with $u \in [0;4]$ and $v \in [0;2]$

So when $u \in [0;2]$ it is necessary that $v<u$ that means to integrate in $int_0^u dv$ and so on...

I other words, if you draw the two lines implied by $0\leq u-v \leq 2$ you get this graph

enter image description here