ELEN90054 Probability and Random Models

Solutions for the Practice Questions on Multivariate Probability Models

questions 1, 2acd, 3ab, 4abcd

1. (a) The total volume under the joint PDF must equal 1. Therefore,
Z ∞Z ∞
1 = fX,Y (x, y) dxdy
−∞ −∞
Z 1Z 1
= kxy dxdy
0 y

1
1 Z 1
kyx2  y ky 3
Z
= dy = − dy
0 2 y 0 2 2
1
ky 2 ky 4  k
= − =
4 8 0 8

This yields k = 8.

(b) From the joint density, we see that the marginal PDF fY (y) is only non-zero for
0 ≤ y < 1. In that region it is calculated as
Z ∞ Z 1 1
8x2 y  8y 8y 3
fY (y) = fX,Y (x, y) dx = 8xy dx = = − .
−∞ y 2 y 2 2

In summary, we write
(
4y − 4y 3 0≤y<1
fY (y) =
0 elsewhere

(c) From the joint density, we see that the marginal PDF fX (x) is non-zero only for
0 ≤ x < 1. In that region the PDF it is calculated as
Z ∞ Z x x
8xy 2 
fX (x) = fX,Y (x, y) dy = 8xy dy = = 4x3
−∞ 0 4 0

In summary, we write (
4x3 0≤x<1
fX (x) =
0 elsewhere

(d) Although this region is a upper quarter-plane, the region of non-zero joint PDF
values is a quadrangle. There could be more

than 1 way to do this, here we integrate
the rectangular region from 21 , 1 , 14 , 21 , followed by the triangle with vertices

1
 1 1
, 1, 12 , (1, 1). Calling the first region A and the second region B, we can



,
2 2
write the probability as (because A and B are disjoint):
  Z Z Z Z
1 1
P X≥ ,Y ≥ = fX,Y (x, y) dxdy + fX,Y (x, y) dxdy
2 4 A B
Z 1Z 1 Z 1Z 1
2
= 8xy dxdy + 8xy dxdy
1 1 1
2
y 4 2

1 1
Z 1 1
Z
2 1
Z 1 Z
2
2 2 3
= 4x y y dy + 4x y  1 dy = 4y − 4y dy + 3y dy
1 1 2 1 1
2 4 2 4

1
1 3y 2  2
2 4 3
= (2y − y ) 1 +
 = (2 − 1) − (1/2 − 1/16) + (1/4 − 1/16)
2 2 1  2
4

27
=
32
2. (a) The joint CDF for any point (x, y) is defined as follow:
Z x Z y
FX,Y (x, y) = P (X ≤ x, Y ≤ y) = fX,Y (u, v) dv du
−∞ −∞

Therefore, we split the regions from 1 up to 9:

For region 1, where (x ≤ 0 or y ≤ 0), obviously we have FX,Y (x, y) = 0;

For region 2, where (0 < x < 1, y ≤ 2),

Z x Z y Z v
y
1 1 
FX,Y (x, y) = dv du = v du
0 6 6 0
0x 0
1  1
= yu = xy
6 0 6

2
For region 3, where (0 < x < 1, y > 2),

Z 2 Z x
1
FX,Y (x, y) = du dv
0 0 6
Z 2
x 2
1  1  1
= u dv = xv  = x
0 6 0 6 0 3

For region 4, where (1 ≤ x < 2, 0 < y ≤ 2),

Z y Z 1
1
FX,Y (x, y) = du dv
0 0 6
Z y
1 y
1  1  1
= u dv = v  = y
0 6 0 6 0 6

For region 5, where (1 < x < 2, y > 2),

Z 2 Z 1
1
FX,Y (x, y) = du dv
0 0 6
Z 2
1 2
1  1  1
= u dv = v  =
0 6 0 6 0 3

3
For region 6, where (2 < x ≤ 4, 0 < y ≤ 2),

Z y Z 1 Z yZ x
1 1
FX,Y (x, y) = du dv + du dv
0 0 6 0 2 6
Z y x
1 1 
= y+ u dv (from region 4)
6 0 6 2
y
1 1 1  1 1 1 1 1
= y + ( x − )v  = y + xy − y = xy − y
6 6 3 0 6 6 3 6 6

For region 7, where (2 < x ≤ 4, y > 2),

Z 2 1
Z 2Z x
Z
1 1
FX,Y (x, y) = du dv + du dv
0 0 6 0 2 6
Z 2 x
1 1 
= + u dv (from region 5)
3 0 6 2
2
1 1 1 
= + ( x − )v 
3 6 3 0
1 1 2 1 1
= + x− = x−
3 3 3 3 3

For region 8, where (x > 4, 0 < y ≤ 2),

4
 Z y Z 1
Z yZ 4
1 1
FX,Y (x, y) = du dv + du dv
0 0 6 0 2 6
Z y 1 Z y 4 y y
1  1  1  1 
= u dv + u dv = v  + v 
0 6 0 0 6 2 6 0 3 0
1 1 1
= y+ y= y
6 3 2

For region 9, where (x > 4, y > 2), FX,Y (x, y) = 1

In summary, we write


 0, x ≤ 0 or y ≤ 0

1
xy, 0 < x ≤ 1, 0 < y ≤ 2


6



1




 3
x, 0 < x ≤ 1, y > 2

1
1 < x ≤ 2, 0 < y ≤ 2



 6
y,

1
FX,Y (x, y) = 3
, 1 < x ≤ 2, y > 2

1
xy − 16 y,




 6
2 < x ≤ 4, 0 < y ≤ 2

1
x − 13 , 2 < x ≤ 4, y > 2


3



1




 2
y, x > 4, 0 < y ≤ 2


1 x > 2, y > 4

(c.) fY (y) is non-zero only for 0 < y < 2, for those values
Z ∞
fY (y) = fX,Y (u, y) du
−∞
Z 1 Z 4 1 4
1 1 1  1 
= du + du =  + 
0 6 2 6 6 0 6 2
1 1 1
= + =
6 3 2
therefore ( 1
2
, 0<y<2
fY (y) =
0, else

5
 (d.) fX (x) is non-zero only for 0 < x < 1 and 2 < x < 4, for those values
Z ∞
fX (x) = fX,Y (x, v) dv
−∞
Z 2 2
1 1  1
= dv =  =
0 6 6 0 3
therefore ( 1
3
, 0 < x < 1, 2 < x < 4
fX (x) =
0, else

3. a. It is the upper-triangular region in the first quadrant with a boundary being the line
y = x.

b. The marginal pdf fX (x) is only non-zero for x ≥ 0. In this region,

Z ∞
fX (x) = fX,Y (x, y) dy
−∞
Z ∞
= 2e−x e−y dy
x
∞
= −2e−x e−y x

= 2e−2x
Thus,
2e−2x , x ≥ 0

fX (x) =
0, elsewhere
Like in the case of fX (x), the marginal pdf fY (y) is only non-zero for y ≥ 0.
Z ∞
fY (y) = fX,Y (x, y) dx
−∞
Z y
= 2e−x e−y dx
0
y
= −2e−x e−y 0

= 2e−y 1 − e−y

Thus,
2e−y (1 − e−y ) , y ≥ 0

fY (y) =
0, elsewhere
Clearly fX,Y (x, y) 6= fX (x)fY (y), so we conclude that the random variables X
and Y are not independent.

4. (a) We take the partial derivative with respect to both x and y.

( y
∂FX,Y (x, y) 2
0 ≤ y < 2, 0 ≤ x < 1
fX,Y (x, y) = =
∂x∂y 0 else

6
(b) From the joint pdf of part a), we see that the marginal pdf fX (x) is only non-zero
for 0 ≤ x < 1. In that region the PDF can be calculated as
Z ∞ Z 2 2
y y 2 
fX,Y (x, y) dy = dy = = 1
−∞ 0 2 4 0

In summary, we write (
1 0≤x<1
fX (x) =
0 else

(c) From the joint pdf of part a), we see that the marginal pdf fY (y) is only non-zero
for 0 ≤ y < 2. In that region the PDF can be calculated as
Z ∞ Z 1
y y
fX,Y (x, y) dx = dx =
−∞ 0 2 2

In summary, we write
y
(
2
0≤y<2
fY (y) =
0 else
(d) The desired probability is over the upper-left region of a quarter-plane starting at
( 12 , 1). However, the joint PDF is only non-zero in the rectangular region defined in
part a). Thus we need only integrate over the region where the PDF is non-zero:
  Z ∞Z 1
1 2
P X≤ ,Y >1 = fX,Y (x, y) dxdy
2 1 −∞
Z 2Z 1
2 y
= dxdy
1 0 2
Z 2  2
1 y 1 y 2 
= · dy =
1 2 2 2 4 1
 
1 1 3
= 1− =
2 4 8