Introductory Probability and Statistical Applications, Second Edition

Paul L. Meyer

Notes and Solutions by David A. Lee

Solutions to Chapter 5: Functions of Random Variables

Suppose that X is uniformly distributed over (−1, 1). Let Y = 4 − X 2 . Find the pdf of Y ,
say g(y), and sketch it. Also verify that g(y) is a pdf.
5.1
By uniform distribution, the pdf of X is f (x) = 1/2, −1 < x < 1. The task is to now find a corresponding pdf
for Y . Given Y = H(X) = 4 − x2 , we can derive the cdf of Y , namely

G(y) = P (Y ≤ y) = P (4 − X 2 ≤ y)
p p
= P (X ≤ − 4 − y, 4 − y ≤ x)
p p
= 1 − P (− 4 − y ≤ x ≤ 4 − y)
Z √4−y √
1  x  4−y
=1− √ dx = 1 − √
− 4−y 2 2 − 4−y
√ √
4−y 4−y p
=− − =− 4−y
2 2
We derive the pdf of Y by differentiating the cdf G(y):

1 1
G0 (y) = g(y) = (4 − y)−1/2 = √
2 2 4−y

Which is distributed over 3 < y < 4, since X is distributed over −1 < x < 1, which maps to 3 < y < 4 under
H(x). To verify g(y) is indeed a pdf, it is clear that g(y) ≥ 0 for the given domain. All that remains to ascertain
R4 1
is whether 3 2√4−y dy = 1. Integrate by u-substitution: let u = 4 − y, du = −1 · dy. Then
Z u=0(y=4) √ 0
1 1
− √ du = − u = 1
2 u=1(y=3) u 1

4
g(y)

0
3 4
y

Suppose that X is uniformly distributed over (1, 3). Obtain the pdf of the following random
variables:
5.2
By uniform distribution, the pdf f (x) = 1/2, 1 < x < 3.

1
 Y = 3X + 4
(a)
Finding the cdf of Y , we get

G(y) = P (Y ≤ y) = P (3X + 4 ≤ y)
 y − 4
=P X≤
3
Z y−4 y−4
3 1 x 3
= dx =
1 2 2 1
y−4 1 y−7
= − =
6 2 6

Differentiating with respect to y, we get G0 (y) = g(y) = 1/6 . Clearly g(y) is positive. For Y = 3X + 4
R 13
such that 1 < x < 3, we can deduce 7 < y < 13. Then 7 1/6 dy = 1 .
1
g(y)

0
7 13
y

Z = eX
(b)
Finding the cdf of Z, we get

G(z) = P (Z ≤ z) = P (eX ≤ z)
= P (X ≤ ln z)
Z ln z
1 x ln z
= dx =
1 2 2 1
ln z 1
= −
2 2
Observe that since the natural logarithm is a strictly increasing function, the inequality direction is pre-
served. Therefore, G0 (z) = g(z) = 1/2z . For Z = eX , 1 < x < 3, we get the domain e < z < e3 for
R e3 3
Z = H(X). Therefore, g(z) ≥ 0 across this domain. Moreover, 21 e z1 dz = 12 ln z|ee = 1 .
1
g(z)

0
e e3
z

2
 Suppose that the continuous random variable X has pdf f (x) = e−x , x > 0. Find the pdf of
the following random variables:
5.3

Y = X3
(a)
The cdf of Y is derived as

G(y) = P (Y ≤ y) = P (X 3 ≤ y)
= P (X ≤ y 1/3 )
Z y1/3 y 1/3 1/3
= e−x dx = −e−x = 1 − e−y
0 0

The pdf of Y then follows:

1 −2/3 −y1/3
G0 (y) = g(y) = y e
3
For Y = X 3 , x > 0, it follows that y > 0. Therefore, g(y) ≥ 0 for y > 0. Numerically evaluating
1 +∞ −2/3 −y 1/3
R
3 0 y e dy gives us 1 .

Z = 3/(X + 1)2
(b)
3
Here, since Z = (X+1)2 is monotonic over the interval 0 < X < +∞, we may apply Theorem 5.1 and find
dx
p p
g(z) = f (x) dz. The inverse function X(z) has two branches, X(z) = 3/z − 1 and X(z) = − 3/z −
p
1. However, because X is distributed over the positive reals, we choose the branch X(z) = 3/z − 1,
1 3 −1/2 1 31/2
defined on 0 < z < 3. Then dx − z32 , therefore dx



dz = 2 z dz = 2 z 3/2 . By Theorem 5.1, g(z) =
31/2 −(√3/z−1) 1 R 3 31/2 −(√3/z−1) 1
e . For 0 < z < 3, g(z) ≥ 0. Moreover, numerical evaluation of 0 2 e z 3/2
dz
2 z 3/2
gives us 1 .

Suppose that the discrete random variable X assumes the values 1, 2, and 3 with equal
probability. Find the probability distribution of Y = 2X + 3.
5.4
Each of the outcomes X = 1, 2, 3 has probability P (X = 1) = P (X = 2) = P (X = 3) = 1/3. Therefore,
H(X = 1) = 5, H(X = 2) = 7, H(X = 3) = 9, and P (Y = 5) = P (Y = 7) = P (Y = 9) = 1/3.

Suppose that X is uniformly distributed over the interval (0, 1). Find the pdf of the following
random variables:
5.5
By uniform distribution, f (x) = 1 for 0 < X < 1.

Y = X2 + 1
(a)
First we find the cdf of Y :

G(y) = P (Y ≤ y) = P (X 2 + 1 ≤ y)
= P (X 2 ≤ y − 1)
p
= P (0 ≤ X ≤ y − 1)
Z √y−1 p
= dx = y − 1
0

Differentiating with respect to y gives us the pdf:

1
G0 (y) = g(y) = (y − 1)−1/2
2

3
 First, note that Y = X 2 + 1, 0 < X < 1 implies 1 < Y < 2. Then g(y) ≤ 0 for 1 < Y < 2. Secondly,
R2 1 2
1 2
(y − 1)−1/2 dy = (y − 1)1/2 1 = 1 .
We may√alternatively apply Theorem 5.1 as Y = X 2 + 1 is monotonic on the interval 0 < X < 1. Then
−1/2
X(y) = y − 1 and dx dx 1
dy = dy = 2 (y − 1) . Therefore g(y) = 12 (y − 1)−1/2 , as expected.

Z = 1/(X + 1)
(b)
First we find the cdf of Z:
 1 
G(z) = P (Z ≤ z) = P ≤z
X +1
1 
=P −1≤X ≤1
z
Z 1
= dx = 1 − (1/z − 1) = 2 − 1/z
1/z−1

Then we derive g(z) as follows:

G0 (z) = g(z) = 1/z 2

Note that Z = 1/(X + 1) over 0 < X < 1 implies 1/2 < Z < 1. Then g(z) ≥ 0 over 1/2 < Z < 1. Moreover,
G(1) − G(1/2) = 1 .
Alternatively, we can apply Theorem 5.1 because Z = 1/(X + 1) is monotonic over 1/2 < Z < 1. Then
X(z) = 1/z − 1 and dx 2 dx 2 2
dz = −1/z , then dz = 1/z . Therefore g(z) = 1/z , as expected.

Suppose that X is uniformly distributed over the interval (−1, 1). Find the pdf of the
following random variables:
5.6
1
By uniform distribution, f (x) = 1−(−1) = 1/2 for −1 < X < 1.

Y = sin (πX/2)
(a)
We first derive the cdf of Y . Here, note that the inverse sine function is increasing over the given interval:

G(y) = P (Y ≤ y) = P (sin(πX/2) ≤ y)
= P (X ≤ (2/π) sin−1 (y))
Z (2/π) sin−1 (y)
1 1 1
= dx = sin−1 (y) +
−1 2 π 2
And now we derive the pdf of Y by differentiating G(y) with respect to y. But first we must determine the
derivative of the inverse sine function.
Theorem.
d 1
sin−1 (x) = √
dx 1 − x2
Proof. First observe that sin(sin−1 (x)) = x. Then dxd
sin(sin−1 (x)) = dx
d
x, and it immediately follows that
−1 d −1 d −1 1
cos(sin (x)) dx sin (x) = 1 =⇒ dx sin (x) = cos(sin−1 (x)) . Now, using the fact that sin2 y + cos2 y = 1,
p p √
and from that deriving cos y = 1 − sin2 y, we can write cos(sin−1 (x)) = 1 − sin2 (sin−1 (x)) = 1 − x2
d
by virtue of inverses. Therefore, dx sin−1 (x) = √1−x
1
2
.
Proceeding, we derive:
d 1 1 1 1
G0 (y) = g(y) = sin−1 (y) + = p
dy π 2 π 1 − y2

For Y = sin(πX/2) on −1 < X < 1, it follows that −1 < Y < 1. Then g(y) ≥ 0 for −1 < Y < 1.
R1
Additionally, −1 π1 √ 1 2 dy = 1 , confirming g(y) is a pdf.
1−y
Alternatively, because Y = sin(πX/2), −1 < X < 1 is monotonic, we may apply Theorem 5.1. Finding
X = (2/π) sin−1 (y) as before, it follows that dx dx
dy = | dy | =
√ 2 2 , and g(y) = 12 · √ 2 2 = √ 1 2 , as
π 1−y π 1−y π 1−y
expected.

4
 Y = cos (πX/2)
(b)
First we derive the cdf of Z. Here, note that the inverse cosine function is strictly decreasing on the given
interval:

G(z) = P (Z ≤ z) = P (cos(πX/2) ≤ z)
 2 2 
= P X ≥ cos−1 (z), X ≤ − cos−1 (z)
π π
Z − π2 cos−1 (z) Z 1
1 1
= dx + dx
−1 2 2
π cos
−1 (z) 2

1 1 1 1
= − cos−1 (z) + − cos−1 (z) +
π 2 π 2
2
= − cos−1 (z) + 1
π
Analogously, we must determine the derivative of the inverse cosine function before proceeding.

Theorem.
d 1
cos−1 (x) = − √
dx 1 − x2

Proof. Since cos(cos−1 (x)) = x, it follows that dx

d
cos(cos−1 (x)) = dx
d
x, implying − sin(cos−1 (x)) dx
d
cos−1 (x) =
p
1. Then dx d
cos−1 (x) = − sin(cos1−1 (x)) . Since sin2 y + cos2 y = 1 implies sin y = 1 − cos2 y, we have
d −1 1
dx cos (x) = − √ 2 −1
= −√ 1 2 .
1−x
1−cos (cos (x))

Differentiating G(z) with respect to z yields the pdf of Z:

d 2  2 1
G0 (z) = g(z) = − cos−1 (z) + 1 = √
dz π π 1 − z2
For Z = cos(πX/2) on −1 < X < 1, the distribution of Z is over 0 < Z < 1. Then g(z) ≥ 0 on that
R1
interval. Moreover, 0 π2 √1−z
1
2
dz = 1 , ascertaining g(z) is a pdf.
Because Z = cos(πX/2) is not monotonic over −1 < X < 1, we cannot apply Theorem 5.1 to derive the
pdf of Z.

W = |X|
(c)
In particular, W is defined as:

W = X, 0≤X<1
n
−X, −1 < X < 0
Beginning with the derivation of the cdf of W :

G(w) = P (W ≤ w) = P (|X| ≤ w)
= P (X ≤ w, −w ≤ X)
Z w Z 0
1 1
= dx + dx
0 2 −w 2
w w
= + =w
2 2
Differentiating with respect to w yields G0 (w) = g(w) = 1 . Since W = |X|, −1 < X < 1 implies
R1
0 < W < 1, clearly g(w) ≥ 0 on that interval. Moreover, 0 w dw = 1 , confirming g(w) is a pdf.

Suppose that the radius of a sphere is a continuous random variable. (Due to inaccuracies
of the manufacturing process, the radii of different spheres may be different.) Suppose that
the radius R has pdf f (r) = 6r(1 − r), 0 < r < 1. Find the pdf of the volume V and the
surface area S of the sphere.
5.7
Volume. The volume of a sphere is V (r) = 43 πr3 . First finding the cdf of the volume v:

5
 4 
G(v) = P (V ≤ v) = P πr3 ≤ v
3
  3 1/3 
=P r≤ v
4π
3 1/3
Z ( 4π v) 3
( 4π v)1/3
= 6r(1 − r) dr = 3r2 − 2r3
0 0
 3 2/3 3
=3 v − v
4π 2π
Differentiating with respect to v gives us the pdf of V :
 3 −1/3  3  3
G0 (v) = g(v) = 2 v −
4π 4π 2π
3  3 −1/3 
= v −1
2π 4π

For V (r) = 43 πr3 on 0 < r < 1, we have 0 < V < 43 π. It follows that g(v) ≥ 0 on this interval, and
R 43 π 3  3 −1/3 
0 2π 4π v − 1 dv = 1 , ascertaining that g(v) is a pdf.
Surface Area. The surface area of a sphere is given by A(r) = 4πr2 . First deriving the cdf of A:

G(a) = P (A ≤ a) = P (4πr2 ≤ a)
  1 1/2 
=P r≤ a
4π
1
Z ( 4π a)1/2 1
( 4π a)1/2
= 6r(1 − r) dr = 3r2 − 2r3
0 0
 1   1 3/2
=3 a −2 a
4π 4π
And now differentiating with respect to a to find the pdf of A:

3  1 1/2  1 
G0 (a) = g(a) = −3 a
4π 4π 4π
3   1  1/2 
= 1− a
4π 4π
R 4π 
3
For A(r) = 4πr2 over 0 < r < 1, we have 0 < A(r) < 4π. Then g(a) ≥ 0 over this interval, and 0 4π 1−
 1/2 
1
4π a da = 1 , ascertaining that g(a) is a pdf.

A fluctuating electric current I may be considered as a uniformly distributed random vari-

able over the interval (9, 11). If this current flows through a 2-ohm resistor, find the pdf of
the power P = 2I 2 .
5.8
1
By uniform distribution, f (i) = 11−9 = 21 , 9 < I < 11, where I is the random variable for current and i a specific
outcome of current. Let P ∗ be the random variable for power and p∗ be a specific outcome of power. Deriving
the cdf of P ∗ gives us:

G(p∗ ) = P (P ∗ ≤ p∗ ) = P (2I 2 ≤ p∗ )
  p∗ 1/2 
=P I≤
2
Z ( p2∗ )1/2 ∗
( p2 )1/2
1 i
= di =
9 2 2 9
1  p∗ 1/2 
= −9
2 2
Deriving the pdf of P ∗ :

6
 1  1  p∗ −1/2  1 
G0 (p∗ ) = g(p∗ ) =
2 2 2 2
∗ −1/2
1 p
 1  2 1/2
= =
8 2 8 p∗

R 242 1  1/2
For P ∗ = 2I 2 , 9 < I < 11, we have 162 < P ∗ < 242. Then g(p∗ ) ≥ 0 and 162 8
2
p∗ dx = 1 , ascertaining
∗
that g(p ) is a pdf.

The speed of a molecule in a uniform gas at equilibrium is a random variable V whose pdf
2
is given by f (v) = av 2 e−bv , v > 0, where b = m/2kT and k, T , and m denote Boltzmann’s
constant, the absolute temperature, and the mass of the molecule, respectively.
5.9

Evaluate the constant a (in terms of b).

(a)
2 2
a −bv
We proceed by integration by parts. Let u = v, du = dv, dw = ave−bv dv, and w = − 2b e . Then
Z +∞ Z +∞
2 a −bv2 +∞ a 2
av 2 e−bv dv = − ve + e−bv dv
0 2b 0 2b 0
r
a 1 π
= =1
2b 2 b
4b3/2
=⇒ a = √
π

Derive the distribution of the random variable W = mv 2 /2, which represents the kinetic
energy of the molecule.
(b)
Without needing to deal with error functions, because W = mv 2 /2 is monotonic for v > 0, we may make
 1/2   −1/2 3/2 2
use of Theorem 5.1. Then V = 2w m , and dv
dw = dv
dw = 1 2
2 m
2w
m . With f (v) = 4b√π v 2 e−bv , we
can derive the pdf of the kinetic energy W :
 
4b3/2  2w  −b 2w
m
 1  2w −1/2
g(w) = √ e
π m m m
2
= w1/2 e−(w/kT ) , W > 0
(kT )3/2 π 1/2

A random voltage X is uniformly distributed over the interval (−k, k). If X is the input of a
nonlinear device with the characteristics shown in Fig. 5.12, find the probability distribution
of Y in the following three cases:
5.10

X
−x0 −a a x0

7
 1
By uniform distribution, f (x) = 2k , −k < X < k.

k<a
(a)
Since the event that Y = 0 is equivalent to the event that X ∈ (−k, k), we may simply calculate P (Y =
Rk 1
0) = −k 2k dx = g(0) = 1 . Therefore, g(y) = 0 , y 6= 0.

a < k < x0
(b)
Here, we define Y piecemeal as follows:
 y0 ay0
 − X− , −k < x < −a
 x0 − a x0 − a


Y = 0, −a ≤ x ≤ a

 y0 ay0

 X− , a<x<k
x0 − a x0 − a
Therefore, we must find the probability distribution function of Y such that
Z y
P (0 ≤ Y ≤ y) = g(y) dy + P (Y = 0) = 1
0
Or equivalently:
Z y(x=−k) Z x=a Z y(x=k)
g(y) dy + f (x) dx + g(y) dy = 1
0(x=−a) x=−a 0(x=a)

First we find G(y) over the X interval (−k, −a):

 y0 ay0 
G(y) = P (Y ≤ y) = P − X− ≤y
x0 − a x0 − a
  x − a  ay0 
0
=P X≥− y+
y0 x0 − a
Z −a
1 a (x0 − a)y a (x0 − a)y
= dx = − + + =
x0 −a
−( y )y−a 2k 2k 2ky0 2k 2ky0
0

x0 − a y0 (k − a)
Then G0 (y) = g(y) = ,0 < y < .
2ky0 x0 − a
Next we find G(y) over the X interval (−a, a). Because Y = 0 over this interval, we may interpret the
events Y = 0 and X ∈ (−a, a) to be equivalent. Then we may simply write
Z a
1 a a a
P (Y = 0) = dx = + =
−a 2k 2k 2k k
Lastly, we drive G(y) over the X interval (a, k):
 y ay0 
0
G(y) = P (Y ≤ y) = P X− ≤y
x0 − a x0 − a
 x − a 
0
=P X≤ y−a
y0
Z ( x0y−a )y−a
0 1 (x0 − a)y a
= dx = −
a 2k 2ky0 k

x0 − a y0 (k − a)
Then G0 (y) = g(y) = ,0 < y < . Since both events X ∈ (−k, −a) and X ∈ (a, k) are
2ky0 x0 − a
equivalent to Y ∈ (0, y), we need only sum the corresponding probability distribution functions of Y over
those respective intervals. In particular, we have:
"Z
−a Z ( x0y−a )y−a #
d d 1 0 1
g(y) = G(y) = dx + dx
dy dy −( x0y−a )y−a 2k a 2k
0

x0 − a y0 (k − a)
= ,0 < y <
ky0 x0 − a

x0 − a y0 (k − a) a
Namely, we can conclude that g(y) = ,0 < y < and g(y) = , y = 0 .
ky0 x0 − a k

8
 k > x0
(c)
By part (b), we know the probability distribution functions over the X range spaces (−a, a), (−x0 , −a),
and (a, x0 ); for the latter two pdfs, the interval over Y for which they are defined is 0 < y < y0 . All that
remains is to determine the probability distribution function of Y for when y = y0 . Simply, because Y is a
constant value for which the equivalent events are X ∈ (−k, −x0 ) and X ∈ (x0 , k), we may write
Z k Z −x0
1 1
P (Y = y0 ) = dx + dx
x0 2k −k 2k
k − x0 k − x0 k − x0
= + = = 1 − x0 /k
2k 2k k
Therefore, the probability distribution function here is:
a/k, y=0


 x0 − a


g(y) = , 0 < y < y0
ky0
x0


1 − , y = y0


k

The radiant energy (in Btu/hr/ft2 ) is given as the following function of temperature T (in
degree Fahrenheit): E = 0.173(T /100)4 . Suppose that the temperature T is considered to
be a continuous random variable with pdf
−2
f (t) = 200t , 40 ≤ t ≤ 50
n
0, elsewhere
Find the pdf of the radiant energy E.
5.11
Let E be the random variable for radiant energy and e a specific outcome of E. First we derive the cdf of E:

G(e) = P (E ≤ e) = P (0.173(t/100)4 ≤ e)
  e 1/4 
= P t ≤ 100
0.173
e
Z 100( 0.173 )1/4  e −3/4
= 200 t−2 dt = −2 +5
40 0.173
Then the pdf of E is:

1  e −3/4  1 
G0 (e) = g(e) =
2 0.173 0.173
 e −5/4
= 2.89 = 0.322e−5/4 , 0.0044 ≤ E ≤ 0.0108
0.173

To measure air velocities, a tube (known as Pitot static tube) is used which enables one to
measure differential pressure. This differential pressure is given by P = 12 dV 2 , where d is
the density of the air and V is the wind speed (mph). If V is a random variable uniformly
distributed over (10, 20), find the pdf of P .
5.12
1
By uniform distribution, f (v) = 10 , 10 < X < 20. We first determine the cdf of X:
1 
G(x) = P (X ≤ x) = P dV 2 ≤ x
2
  2X 1/2 
=P V ≤
d
Z ( 2X 1/2
d ) 1 1  2X 1/2
= dv = −1
10 10 10 d
The pdf of X is then:

1  2  2X −1/2 1  2X −1/2
G0 (x) = g(x) = = , 50d < X < 200d
20 d d 10d d

9
 R 200d  −1/2
1 2X
Which is clearly positive for X ∈ (50d, 200d), and it also follows that 50d 10d d dx = 1 .

Suppose that P (X ≤ 0.29) = 0.75, where X is a continuous random variable with some
distribution defined over (0, 1). If Y = 1 − X, determine k so that P (Y ≤ k) = 0.25.
5.13
R 0.29 R1
By premise, P (X ≤ 0.29) = 0 f (x) dx = 0.75. Since it must be the case that P (0 ≤ X ≤ 1) = 0 f (x) dx = 1,
R1
it immediately follows that 0.29 f (x) dx = 0.25. Lastly, in finding the cdf of Y , we determine

G(y) = P (Y ≤ y) = P (1 − X ≤ y)
= P (X ≥ 1 − y)
Z 1
= f (x) dx
1−y
R1
Now, suppose y = k, then 1−y f (x) dx = 0.25. Therefore, 1−y = 0.29 and y = 0.71 . Observe that determining
the pdf of X, f (x), was completely unnecessary.

10