If X and Y are independent and identically distributed uniform random variables on (0,1), compute each of the following joint densities. (a) U=5X, V=4X/Y. fU,V(u,v)= (b) U=3X+Y, V=5X/(X+Y). fU,V(u,v)=

Respuesta :

Use the method of transformations for both exercises.

a.

[tex]\begin{cases}U=5X\\V=\frac{4X}Y\end{cases}\implies\begin{cases}X=\frac U5\\Y=\frac{4U}{5V}\end{cases}[/tex]

The Jacobian for this transformation is

[tex]J=\begin{bmatrix}X_U&X_V\\Y_U&Y_V\end{bmatrix}=\begin{bmatrix}\frac15&0\\\frac4{5V}&-\frac{4U}{5V^2}\end{bmatrix}\implies|\det J|=\frac{4U}{25V^2}[/tex]

Then the joint density of [tex]U[/tex] and [tex]V[/tex] is

[tex]f_{U,V}(u,v)=f_{X,Y}\left(\dfrac U5,\dfrac{4U}{5V}\right)|\det J|[/tex]

Both [tex]X[/tex] and [tex]Y[/tex] are i.i.d with density

[tex]f_R(r)=\begin{cases}1&\text{for }0<r<1\\0&\text{otherwise}\end{cases}[/tex]

(where [tex]R[/tex] is either [tex]X[/tex] or [tex]Y[/tex]) so we have

[tex]f_{U,V}(u,v)=\begin{cases}\frac{4u}{25v^2}&\text{for }0<\frac u5<1\text{ and }0<\frac{4u}{5v}<1\\0&\text{otherwise}\end{cases}[/tex]

We can clean up the support a bit:

[tex]0<\dfrac{4u}{5v}<1\implies\dfrac{5v}{4u}<0\text{ or }\dfrac{5v}{4u}>1[/tex]

The first case suggests either [tex]U[/tex] or [tex]V[/tex] can be negative, but with [tex]U=5X[/tex], that is impossible. This leaves us with

[tex]\dfrac{5v}{4u}>1\implies v>\dfrac{4u}5[/tex]

so the joint PDF is

[tex]f_{U,V}(u,v)=\begin{cases}\frac{4u}{25v^2}&\text{for }0<\frac u5<1\text{ and }v>\frac{4u}5\\0&\text{otherwise}\end{cases}[/tex]

b.

[tex]\begin{cases}U=3X+Y\\V=\frac{5X}{X+Y}\end{cases}\implies\begin{cases}X=\frac{UV}{5+2V}\\Y=\frac{(5-V)U}{5+2V}\end{cases}[/tex]

The Jacobian is

[tex]J=\begin{bmatrix}\frac V{5+2V}&\frac{5U}{(5+2V)^2}\\\frac{5-V}{5+2V}&-\frac{15UV}{(5+2V)^2}\end{bmatrix}\implies|\det J|=\dfrac{5U}{(5+2V)^2}[/tex]

and so the joint PDF is

[tex]f_{U,V}(u,v)=f_{X,Y}\left(\frac{uv}{5+2v},\frac{(5-v)u}{5+2v}\right)|\det J|[/tex]

[tex]f_{U,V}(u,v)=\begin{cases}\frac{5u}{(5+2v)^2}&\text{for }0<\frac{uv}{5+2v}<1,0<\frac{(5-v)u}{5+2v}<1\\0&\text{otherwise}\end{cases}[/tex]

We have

[tex]0<\dfrac{uv}{5+2v}<1\implies0<uv<5+2v[/tex]

and

[tex]0<\dfrac{(5-v)u}{5+2v}<1\implies0<5u-uv<5+2v[/tex]

which together tell us that

[tex]0<5u<2(5+2v)\implies0<u<2+\dfrac{4v}5[/tex]

As for [tex]v[/tex], we know that at most we can have [tex]X=Y=1[/tex], so that [tex]V[/tex] can have a maximum value of 5/2, so we can simplify the support to write the PDF as

[tex]f_{U,V}(u,v)=\begin{cases}\frac{5u}{(5+2v)^2}&\text{for }0<u<2+\frac{4v}5,0<v<\frac52\\0&\text{otherwise}\end{cases}[/tex]