Math 3215 Intro.

Probability & Statistics Summer ’14

Homework 5: Due 7/3/14

1. Let X and Y be continuous random variables with joint/marginal p.d.f.’s

f (x, y) = 2, 0 ≤ x ≤ y ≤ 1,
f1 (x) = 2(1 − x), 0 ≤ x ≤ 1,
f2 (y) = 2y, 0 ≤ y ≤ 1.
1 3

 X=1 .


Find the conditional p.d.f. h(y|x) and the conditional probability P 2
≤Y ≤ 4 4
What is the expected value of Y when X = 14 ?
Solution: The conditional p.d.f h(y|x) = f (x, y)/f1 (x) is immediately seen to be
2 1
h(y|x) = = .
2(1 − x) 1−x

To find P ( 21 ≤ Y ≤ 34 | X = 14 ) we integrate the conditional p.d.f. h(y| 41 ) on the interval

1/2 ≤ y ≤ 3/4, and we obtain
   Z 3/4  
1 3  1 1 1 1 1
P ≤Y ≤  X= = 1 dy = 3 = .
2 4 4 1/2 1 − 4 4 4 3

Since expectation is linear, we have E(Y |X = 14 ) = E(4/3) = 4/3 .

2. Let X and Y be discrete random variables with joint p.m.f.

x+y
f (x, y) = , x = 1, 2, y = 1, 2, 3, 4.
32
Find the marginal p.m.f.’s of X and Y and the conditional p.m.f.’s g(x|y) and h(y|x). Find
P (1 ≤ Y ≤ 3 | X = 1) and P (Y ≤ 2 | X = 2). Finally, find E(Y | X = 1) and find
V ar(Y | X = 1).
Solution: The marginal p.m.f.’s of X and Y are immediately seen to be
X 4x + 10
f1 (x) = f (x, y) = ,
y=1,2,3,4
32
X 3 + 2y
f2 (y) = f (x, y) = .
x=1,2
32
 2

The conditional p.m.f.’s are thus seen to be

f (x, y) (x + y)/32 x+y

h(y|x) = = = ,
f1 (x) (4x + 10)/32 4x + 10
f (x, y) (x + y)/32 x+y
g(x|y) = = = .
f2 (y) (3 + 2y)/32 3 + 2y

Also, we have
X 2 3 4 9
P (1 ≤ Y ≤ 3 | X = 1) = h(y|1) = + + = .
y=1,2,3
14 14 14 14

5 9
Note, this can also be computed as 1 − h(4|1) = 1 − 14
= 14
. Next, we compute

X 2+1 2+2 7
P (Y ≤ 2 | X = 2) = h(y|2) = + = .
y=1,2
18 18 18

Finally, we have that

X 2 3 4 4 18
E(Y | X = 1) = y · h(y|1) = (1) + (2) + (3) + (4) = ,
y=1,2,3,4
14 14 14 14 7

and  2
2 57 18
V ar(Y | X = 1) = E(Y | X = 1) − E(Y | X = 1) = − = 1.503 .
7 7

2
 3

3. Let W equal the weight of a box of oranges which is supposed to weight 1-kg. Suppose that
P (W < 1) = .05 and P (W > 1.05) = .1. Call a box of oranges light, good, or heavy depending
on if W < 1, 1 ≤ W ≤ 1.05, or W > 1.05, respectively. In n = 50 independent observations
of these boxes, let X equal the number of light boxes and Y the number of good boxes.
Find the joint p.m.f. of X and Y . How is Y distributed? Name the distribution and state the
values of the parameters associated to this distribution. Given X = 3, how is Y distributed?
Determine E(Y | X = 3) and find the correlation coefficient ρ of X and Y .
Solution: The random variables are said to come from a trinomial distribution in this case
since there are three exhaustive and mutually exclusive outcomes light, good, or heavy, having
probabilities p1 = .05, p2 = .85, and p3 = .1, respectively. It is easy to see that the trinomial
p.m.f. in this case is
50!
f (x, y) = px py p50−x−y .
x!y!(50 − x − y)! 1 2 3
That is, f (x, y) is exactly the joint p.m.f. of X and Y , the number of light boxes and good
boxes, where p1 = .05, p2 = .85, and p3 = .1 are the various probabilities of each of the three
events: light, good, and heavy.
The random variable is binomially distributed, but the parameters of the distribution depend
on the value of the random variable X. We have that the random variable Y is (conditionally)
p2
binomially distributed b(n − x, 1−p 1
) since the marginal distributions of X, Y are b(n, p1 ),
b(n, p2 ) and the conditional p.m.f. of Y is thus, with n = 50,

1 n! x!(n − x)! 1
h(y|x) = f (x, y) = px1 py2 pn−x−y
3 · x
f1 (x) x!y!(n − x − y)! n! p1 (1 − p1 )n−x
n!x!(n − x)! px1 py2 pn−x−y
3
= · x· ·
x!y!(n − x − y)!n! p1 (1 − p1 ) (1 − p1 ) (1 − p1 )−y
y n−x
 y  n−x−y
(n − x)! p2 p3
= .
y!(n − x − y)! 1 − p1 1 − p1
In this case, with the specified values of n, p1 , p2 , and p3 , we have for X = 3
47!
h(y|3) = (.8947)y (.1053)47−y ,
y!(47 − y)!

so that Y is conditionally b(47, .8947) when X = 3. Since µ = np for binomial distribution,

we have that E(Y |X = 3) = (47)(.8947) = 42.05 .
p2
It is not hard to see that in fact E(Y |x) = (n − x) 1−p 1
in general, and that a similar formula
holds for E(X|y). The correlation coefficient is now found using the fact that since each of
 4

p2 p1
the conditional expectations E(Y |x) = (n − x) 1−p 1
and E(X|y) = (n − y) 1−p 2
is linear, then
2
the square of the correlation coefficient ρ is equal to the product of the respective coefficients
of x and y in the conditional expectations.
  
2 −p2 −p1 p1 p2
ρ = = ,
1 − p1 1 − p2 (1 − p1 )(1 − p2 )

from which it follows that

r
p1 p 2
ρ= − = −.0819 .
(1 − p1 )(1 − p2 )

The fact that the correlation coefficient is negative follows from the fact that, for example,
σY
E(Y |x) = µY + ρ (x − µX ),
σX
and noting that the coefficient of x in E(Y |x) is seen to be negative (and also σY , σX > 0).

4. Let X have the uniform distribution U (0, 2) and let the conditional distribution of Y , given
that X = x, be U (0, x). Find the joint p.d.f. f (x, y) of X and Y , and be sure to state the
domain of f (x, y). Find E(Y |x).
Solution: We have that
y
f (x, y) = , 0 < x ≤ 2, 0 ≤ y ≤ x.
x
Now, x
x
x2
Z
y 1 3 
E(Y |x) = y · dx = y  = .
0 x 3x 0 3

2
 5

5. The support of a random variable X is the set of x-values such that f (x) 6= 0. Given that
X has p.d.f. f (x) = x2 /3, −1 < x < 2, what is the support of X 2 ? Find the p.m.f. of the
random variable Y = X 2 .
Solution: The p.d.f. g(y) of Y = X 2 is obtained as follows. We note that the possible y-values
that can be obtained are in the range 0 ≤ y ≤ 4, so the support of g(y) needs to be the interval
√
[0, 4]. Now, on the interval [1, 4], there is a one-to-one transformation represented by x = y.
√
We first find G(y) = P (Y ≤ y) = P (X 2 ≤ y) = P (X ≤ y) for X = x in [1, 2], corresponding
to Y = y in [1, 4]. We have
√ √
y y
x2
Z Z
G(y) = f (x) dx = dx,
1 1 3
√ √
( y)2
and in particular g(y) = G0 (y) = · ( y)0 , from
3 √
the chain rule and the Fundamental
y
Theorem of Calculus. Simplifying, we have g(y) = 6 , 1 < y < 4.
In order to find g(y) on 0 < y < 1, we need to work a little harder. For x in the interval −1 <
√ √
x < 1 there is a two-to-one transformation given by x = − y for −1 < x < 0, and x = y
for 0 < x < 1. We then calculate G(y) for 0 < y < 1 (i.e., −1 < x < 1) as before, but now
√ √
using two integrals G(y) = P (Y ≤ y) = P (X 2 ≤ y) = P (− y < X < 0) + P (0 < X < y),
so for y in the interval 0 < y < 1 we have
Z 0 Z √y
G(y) = √ f (x) dx + f (x) dx.
− y 0

Again, from the chain rule and the Fundamental Theorem of Calculus we have
√
0 √ √ 0 √ √ 0 −y −1 y 1 y
g(y) = G (y) = −f (− y) · (− y) + f ( y) · ( y) = · √ + · √ = .
3 2 y 3 2 y 3

So,
(√
y/3 if 0 < y < 1,
g(y) = √
y/6 if 1 < y < 4.

There is no problem defining g(0) = g(1) = 0, or even just leaving the p.d.f. undefined at the
points y = 0 and y = 1.

2
 6

6. Let X1 , X2 denote two independent random variables each with the χ2 (2) distribution. Find
the joint p.d.f. of Y1 = X1 and Y2 = X1 + X2 . What is the support of Y1 , Y2 (i.e., what is the
domain of the joint p.d.f., where f (y1 , y2 ) 6= 0)? Are Y1 and Y2 independent?
Solution: We have that X1 , X2 have the same p.d.f.
1 −x/2
h(x) =
e , 0 ≤ x < ∞,
2
corresponding to the χ2 (r) distribution with r = 2 degrees of freedom. By the way, this is
the same as saying that X1 , X2 follow exponential distributions with θ = 2. Since X1 , X2 are
independent, the joint p.d.f. of X1 and X2 is
1 −(x1 +x2 )/2
f (x1 , x2 ) = h(x1 )h(x2 ) = e .
4
The change of variables formula is g(y1 , y2 ) = |J|f (v1 (y1 , y2 ), v2 (y1 , y2 )) using the determinant
of the Jacobian  
 ∂v1 (x1 , x2 ) ∂v1 (x1 , x2 ) 
 

 ∂y 1 ∂y 2


J =  

 ∂v2 (x1 , x2 ) ∂v2 (x1 , x2 ) 
 
 ∂y1 ∂y2 
where vi (y1 , y2 ) is the inverse of ui , Yi = ui (X1 , X2 ), i = 1, 2. In this case, Y1 = u1 (X1 , X2 ) =
X1 , so X1 = v1 (Y1 , Y2 ) = Y1 , and Y2 = u2 (X1 , X2 ) = X1 + X2 , so X2 = v2 (Y1 , Y2 ) = Y2 − Y1 .
Hence,  
 1 0
|J| =   = 1,
−1 1
so the joint p.d.f. of Y1 and Y2 is
1
g(y1 , y2 ) = |1|f (y1 , y2 − y1 ) = e−y2 /2 , 0 ≤ y1 ≤ y2 < ∞.
4
To determine if Y1 and Y2 are independent we compute the marginal p.d.f.’s. We have,
Z y2
1 −y2 /2 y2
g1 (y1 ) = e dy1 = e−y2 /2 ,
0 4 4
and ∞
∞
−1 −y2 /2 
Z
1 −y2 /2 1
g2 (y2 ) = e dy2 = e = e−y1 /2 .
y1 4 2 
y1 2
Since, g(y1 , y2 ) 6= g1 (y1 )g2 (y2 ), Y1 and Y2 are dependent.