The Distribution of a Linear Combination
5.5 The Distribution of a Linear Combination
The sample mean X and sample total T o are special cases of a type of random vari- able that arises very frequently in statistical applications.
DEFINITION Given a collection of n random variables X 1 ,...,X n and n numerical constants
a 1 ,...,a n , the rv
n
Y5a 1 X 1 1 …1 a n X n 5 o a i X i (5.7)
i5 1
is called a linear combination of the X i ’s.
For example, consider someone who owns 100 shares of stock A, 200 shares of stock
B, and 500 shares of stock C. Denote the share prices of these three stocks at some
particular time by X 1 ,X 2 , and X 3 , respectively. Then the value of this individual’s stock holdings is the linear combination Y 5 100X 1 1 200X 2 1 500X 3 . Taking a 1 5a 2 5...5a n 5 1 gives Y 5 X 1 1...1X n 5T o , and a 1 5
a 2 5…5 a n 5 1 yn yields
Notice that we are not requiring the X i ’s to be independent or identically distrib- uted. All the X i ’s could have different distributions and therefore different mean values and variances. Our first result concerns the expected value and variance of
a linear combination.
prOpOSITION Let X 1 ,X 2 ,...,X n have mean values m 1 ,...,m n , respectively, and variances
s 2 1 2 , …, s n , respectively.
1. Whether or not the X i ’s are independent,
E (a 1 X 1 1a 2 X 2 1...1a n X n )5a 1 E (X 1 )1a 2 E (X 2 )1...1a n E (X n )
5 a 1 m 1 1...1a n m n (5.8)
5.5 the Distribution of a Linear Combination 239
2. If X 1 ,...,X n are independent,
a 1 X 1 1…1 a n X n 5 Ïa 1 s 1 1…1 a n s n (5.10)
3. For any X 1 ,...,X n ,
i5 o 1 j5 o 1
a i a j Cov(X i ,X j ) (5.11)
Proofs are sketched out at the end of the section. A paraphrase of (5.8) is that the expected value of a linear combination is the same as the linear combination of
the expected values—for example, E(2X 1 1 5X 2 ) 5 2m 1 1 5m 2 . The result (5.9) in
Statement 2 is a special case of (5.11) in Statement 3; when the X i ’s are independent, Cov(X i , X j ) 5 0 for i ? j and 5 V(X i ) for i 5 j (this simplification actually occurs when the X i ’s are uncorrelated, a weaker condition than independence). Specializing to the case of a random sample (X i ’s iid) with a i 5 1n for every i gives E(X) 5 m and
V (X) 5 s 2 n, as discussed in Section 5.4. A similar comment applies to the rules for T o . ExamplE 5.30 A gas station sells three grades of gasoline: regular, extra, and super. These are
priced at 3.00, 3.20, and 3.40 per gallon, respectively. Let X 1 ,X 2 , and X 3 denote
the amounts of these grades purchased (gallons) on a particular day. Suppose the
X i ’s are independent with m 1 5 1000, m 2 5 500, m 3 5 300, s 1 5 100, s 2 5 80, and s 3 5 50. The revenue from sales is Y 5 3.0X 1 1 3.2X 2 1 3.4X 3 , and
E (Y) 5 3.0m 1 1 3.2m 2 1 3.4m 3 5 5620
V (Y) 5 (3.0) 2 s 2 1 1 (3.2) 2 s 2 1 (3.4) 2 s 2 3 5 184,436
s Y 5 Ï184,436 5 429.46
n
the difference Between two random Variables
An important special case of a linear combination results from taking n 5 2, a 1 5 1,
and a 2 5 21: Y 5a 1 X 1 1a 2 X 2 5X 1 2X 2 We then have the following corollary to the proposition.
COrOllarY E (X 1 2X 2 ) 5 E(X 1 ) 2 E(X 2 ) for any two rv’s X 1 and X 2 .
V (X 1 2X 2 ) 5 V(X 1 ) 1 V(X 2 ) if X 1 and X 2 are independent rv’s.
The expected value of a difference is the difference of the two expected values. However, the variance of a difference between two independent variables is the sum,
not the difference, of the two variances. There is just as much variability in X 1 2X 2 as in X 1 1X 2 [writing X 1 2X 2 5X 1 1 (21)X 2 , (21)X 2 has the same amount of
variability as X 2 itself].
240 ChapteR 5 Joint probability Distributions and Random Samples
ExamplE 5.31 A certain automobile manufacturer equips a particular model with either a six-cylinder
engine or a four-cylinder engine. Let X 1 and X 2 be fuel efficiencies for independently and randomly selected six-cylinder and four-cylinder cars, respectively. With m 1 5 22,
m 2 5 26, s 1 5 1.2, and s 2 5 1.5,
E (X 1 2X 2 )5m 1 2m 2 5 22 2 26 5 24
V (X 1 2 X 2 )5s 2 1 1s 2 5 (1.2) 2 1 (1.5) 2 5 3.69
s X 1 2 X 2 5 Ï3.69 5 1.92
If we relabel so that X 1 refers to the four-cylinder car, then E(X 1 2X 2 ) 5 4, but the
variance of the difference is still 3.69.
n
the case of normal random Variables
When the X i ’s form a random sample from a normal distribution, X and T o are both normally distributed. Here is a more general result concerning linear combinations.
prOpOSITION If X 1 ,X 2 ,…, X n are independent, normally distributed rv’s (with possibly dif- ferent means andor variances), then any linear combination of the X i ’s also
has a normal distribution. In particular, the difference X 1 2X 2 between two
independent, normally distributed variables is itself normally distributed.
ExamplE 5.32 The total revenue from the sale of the three grades of gasoline on a particular day
(Example 5.30
was Y 5 3.0X 1 1 3.2X 2 1 3.4X 3 , and we calculated m Y 5 5620 and (assuming
continued) independence) s Y 5 429.46. If the X i ’s are normally distributed, the probability that
revenue exceeds 4500 is
P(Y . 4500) 5 P Z.
5 P(Z . 22.61) 5 1 2 F(22.61) 5 .9955
n
The CLT can also be generalized so it applies to certain linear combinations. Roughly speaking, if n is large and no individual term is likely to contribute too much to the overall value, then Y has approximately a normal distribution.
Proofs for the Case n52
For the result concerning expected values, suppose that X 1 and X 2 are continuous
with joint pdf f(x 1 ,x 2 ). Then
sa 1
1 X 1 sx 1 d dx 1
5a 1 x f
a 2 x
f X 2 sx 2 d dx 2
5 a 1 E (X 1 )1a 2 E (X 2 )
5.5 the Distribution of a Linear Combination 241
Summation replaces integration in the discrete case. The argument for the variance result does not require specifying whether either variable is discrete or continuous.
Recalling that V(Y) 5 E[(Y 2 m Y ) 2 ],
V (a 1 X 1 1a 2 X 2 ) 5 E{[a 1 X 1 1a 2 X 2 2 (a 1 m 1 1a 2 m 2 )] 2 }
5 E {a 1 2 sX 1 2m 1 d 2 1 a 2 sX 2 2m 2 d 2 1 2a 1 a 2 sX 1 2m 1 dsX 2 2m 2 d} The expression inside the braces is a linear combination of the variables Y 1 5
(X 1 2m 1 ) 2 , Y 2 5 (X 2 2m 2 ) 2 , and Y 3 5 (X 1 2m 1 )(X 2 2m 2 ), so carrying the E operation through to the three terms gives a 1 2 V sX 1 d1a 2 V sX 2 d 1 2a 1 a 2 Cov sX 1 ,X 2 d
as required.
n
EXERCISES Section 5.5 (58–74)
58. A shipping company handles containers in three different
61. Exercise 26 introduced random variables X and Y, the
sizes: (1) 27 ft 3 (3 3 3 3 3), (2) 125 ft 3 , and (3) 512 ft 3 .
number of cars and buses, respectively, carried by a
Let X i (i 5 1, 2, 3) denote the number of type i containers
ferry on a single trip. The joint pmf of X and Y is given
shipped during a given week. With m i 5 E(X i ) and
in the table in Exercise 7. It is readily verified that X
s i 2 5 V (X i ), suppose that the mean values and standard
and Y are independent.
deviations are as follows:
a. Compute the expected value, variance, and standard
m 1 5 200 m 2 5 250 m 3 5 100
deviation of the total number of vehicles on a single trip.
s 1 5 10 s 2 5 12 s 3 58
b. If each car is charged 3 and each bus 10, compute
a. Assuming that X 1 ,X 2 ,X 3 are independent, calculate
the expected value, variance, and standard deviation
the expected value and variance of the total volume
of the revenue resulting from a single trip.
shipped. [Hint: Volume 5 27X 1 1 125X 2 1 512X 3 .]
b. Would your calculations necessarily be correct if the
62. Manufacture of a certain component requires three
X i ’ s were not independent? Explain.
different machining operations. Machining time for each operation has a normal distribution, and the three
59. Let X 1 ,X 2 , and X 3 represent the times necessary to per-
times are independent of one another. The mean values
form three successive repair tasks at a certain service
are 15, 30, and 20 min, respectively, and the standard
facility. Suppose they are independent, normal rv’s with
deviations are 1, 2, and 1.5 min, respectively. What is
expected values m 1 ,m 2 , and m 3 and variances s 1 2 ,s 2 , and
the probability that it takes at most 1 hour of machining
s 3 2 , respectively.
time to produce a randomly selected component?
a. If m 5m 2 5m 3 5 60 and s 1 2 5s 2 5s 3 2 5 15, calcu-
late P( T o 200) and P(150 T o 200)?
63. Refer to Exercise 3.
b. Using the m i ’ s and s
i ’ s given in part (a), calculate
a. Calculate the covariance between X
1 5 the number
both P(55 X) and P(58 X 62).
of customers in the express checkout and X 2 5 the
c. Using the m i ’ s and s i ’ s given in part (a), calculate
number of customers in the superexpress checkout.
and interpret P(210 X 1 2 .5X 2 2 .5X
b. Calculate V(X
1 1X 2 ). How does this compare to
and s 3 2 5 14, calculate P(X 1 1X 2 1X 3 160) and
64. Suppose your waiting time for a bus in the morning is
also P(X 1 1X 2 2X 3 ).
uniformly distributed on [0, 8], whereas waiting time in
60. Refer back to Example 5.31. Two cars with six-cylinder
the evening is uniformly distributed on [0, 10] indepen-
engines and three with four-cylinder engines are to be driv-
dent of morning waiting time.
en over a 300-mile course. Let X 1 ,...X 5 denote the resulting
a. If you take the bus each morning and evening for a
fuel efficiencies (mpg). Consider the linear combination
week, what is your total expected waiting time? [Hint: Define rv’s X , ..., X and use a rule of
Y 5 (X 1 1X 2 )
y2 2 (X 10
3 1X 4 1 1 X 5 ) y3 expected value.]
which is a measure of the difference between four-
b. What is the variance of your total waiting time?
cylinder and six-cylinder vehicles. Compute P(0 Y)
c. What are the expected value and variance of the dif-
and P(Y . 22). [Hint: Y 5 a 1 X 1 1 ... 1a 5 X 5 , with
ference between morning and evening waiting times
a 1 5 1 y2 ,…, a 5 52 1 y3.]
on a given day?
242 ChapteR 5 Joint probability Distributions and Random Samples
d. What are the expected value and variance of the dif-
value 1 in. and standard deviation .1 in. Assuming that
ference between total morning waiting time and total
the lengths and amount of overlap are independent of one
evening waiting time for a particular week?
another, what is the probability that the total length after
65. Suppose that when the pH of a certain chemical
insertion is between 34.5 in. and 35 in.?
compound is 5.00, the pH measured by a randomly
68. Two airplanes are flying in the same direction in adjacent
selected beginning chemistry student is a random vari-
parallel corridors. At time t 5 0, the first airplane is
able with mean 5.00 and standard deviation .2. A large
10 km ahead of the second one. Suppose the speed of the
batch of the compound is subdivided and a sample
first plane (kmhr) is normally distributed with mean 520
given to each student in a morning lab and each student
and standard deviation 10 and the second plane’s speed
in an afternoon lab. Let X 5 the average pH as deter-
is also normally distributed with mean and standard
mined by the morning students and Y 5 the average pH
deviation 500 and 10, respectively.
as determined by the afternoon students.
a. What is the probability that after 2 hr of flying, the
a. If pH is a normal variable and there are 25 students
second plane has not caught up to the first plane?
in each lab, compute P(2.1 X 2 Y .1). [Hint:
b. Determine the probability that the planes are sepa-
X 2 Y is a linear combination of normal variables,
rated by at most 10 km after 2 hr.
so is normally distributed. Compute m X2Y and s X2Y .]
69. Three different roads feed into a particular freeway
b. If there are 36 students in each lab, but pH determi-
entrance. Suppose that during a fixed time period, the
nations are not assumed normal, calculate (approxi-
number of cars coming from each road onto the freeway
mately) P(2.1 X 2 Y .1).
is a random variable, with expected value and standard
66. If two loads are applied to a cantilever beam as shown in
deviation as given in the table.
the accompanying drawing, the bending moment at 0 due
Road 1
Road 2 Road 3
to the loads is a 1 X 1 1a 2 X 2 .
Expected value
X 1 X 2 Standard deviation
16 25 18 a. What is the expected total number of cars entering
a 1 a 2 the freeway at this point during the period? [Hint: 0 Let X i 5 the number from road i.]
b. What is the variance of the total number of entering
a. Suppose that X 1 and X 2 are independent rv’s with
cars? Have you made any assumptions about the
means 2 and 4 kips, respectively, and standard devia-
relationship between the numbers of cars on the
tions .5 and 1.0 kip, respectively. If a 1 5 5 ft and
different roads?
a 2 5 10 ft, what is the expected bending moment and
c. With X i denoting the number of cars entering from
what is the standard deviation of the bending
road i during the period, suppose that Cov(X 1 ,X 2 ) 5 80,
moment?
Cov(X 1 ,X 3 ) 5 90, and Cov(X 2 ,X 3 ) 5 100 (so that the
b. If X 1 and X 2 are normally distributed, what is the
three streams of traffic are not independent). Compute
probability that the bending moment will exceed
the expected total number of entering cars and the
75 kip-ft?
standard deviation of the total.
c. Suppose the positions of the two loads are random
70. Consider a random sample of size n from a continuous
variables. Denoting them by A 1 and A 2 , assume that
distribution having median 0 so that the probability of
these variables have means of 5 and 10 ft, respec-
any one observation being positive is .5. Disregarding the
tively, that each has a standard deviation of .5, and
signs of the observations, rank them from smallest to
that all A i ’s and X i ’s are independent of one another.
largest in absolute value, and let W 5 the sum of the
What is the expected moment now?
ranks of the observations having positive signs. For
d. For the situation of part (c), what is the variance of
example, if the observations are 2.3, 1.7, 12.1, and
the bending moment?
2 2.5, then the ranks of positive observations are 2 and 3,
e. If the situation is as described in part (a) except that
so W 5 5. In Chapter 15, W will be called Wilcoxon’s
Corr(X 1 ,X 2 ) 5 .5 (so that the two loads are not inde-
signed-rank statistic. W can be represented as follows:
pendent), what is the variance of the bending moment?
W 51?Y 1 12?Y 2 13?Y 3 1...1n?Y n
67. One piece of PVC pipe is to be inserted inside another n
o i?Y i i5 1
piece. The length of the first piece is normally distributed
with mean value 20 in. and standard deviation .5 in. The length of the second piece is a normal rv with mean and
where the Y i ’ s are independent Bernoulli rv’s, each with
standard deviation 15 in. and .4 in., respectively. The
p 5 .5 (Y i 5 1 corresponds to the observation with rank
amount of overlap is normally distributed with mean
i being positive).
Supplementary exercises 243
a. Determine E(Y i ) and then E(W) using the equation
s 3 5 2, m 4 5 12, s 4 5 3. I plan to leave my office at
for W. [Hint: The first n positive integers sum to
precisely 10:00 a.m. and wish to post a note on my door
n (n 1 1) y 2.]
that reads, “I will return by t a.m.” What time t should I
b. Determine V(Y i ) and then V(W). [Hint: The sum of
write down if I want the probability of my arriving after
the squares of the first n positive integers can be
t to be .01?
expressed as n(n 1 1)(2n 1 1) y 6.]
73. Suppose the expected tensile strength of type-A steel is
71. In Exercise 66, the weight of the beam itself contributes
105 ksi and the standard deviation of tensile strength is
to the bending moment. Assume that the beam is of uni-
8 ksi. For type-B steel, suppose the expected tensile
form thickness and density so that the resulting load is
strength and standard deviation of tensile strength are
uniformly distributed on the beam. If the weight of the
100 ksi and 6 ksi, respectively. Let X 5 the sample aver-
beam is random, the resulting load from the weight is
age tensile strength of a random sample of 40 type-A
also random; denote this load by W (kip-ft).
specimens, and let Y 5 the sample average tensile
a. If the beam is 12 ft long, W has mean 1.5 and stan-
strength of a random sample of 35 type-B specimens.
dard deviation .25, and the fixed loads are as described
a. What is the approximate distribution of X? Of Y?
in part (a) of Exercise 66, what are the expected value
b. What is the approximate distribution of X 2 Y?
and variance of the bending moment? [Hint: If the
Justify your answer.
load due to the beam were w kip-ft, the contribution
c. Calculate (approximately) P(21 X 2 Y 1).
to the bending moment would be w 0 12 x dx.]
d. Calculate P(X 2 Y 10). If you actually observed
b. If all three variables (X 1 , X 2 , and W) are normally
X 2 Y 10, would you doubt that m 1 2m 2 5 5?
distributed, what is the probability that the bending
74. In an area having sandy soil, 50 small trees of a certain
moment will be at most 200 kip-ft?
type were planted, and another 50 trees were planted in an
72. I have three errands to take care of in the Administration
area having clay soil. Let X 5 the number of trees planted
Building. Let X i 5 the time that it takes for the ith errand
in sandy soil that survive 1 year and Y 5 the number of
(i 5 1, 2, 3), and let X 4 5 the total time in minutes that
trees planted in clay soil that survive 1 year. If the prob-
I spend walking to and from the building and between
ability that a tree planted in sandy soil will survive 1 year
each errand. Suppose the X i ’ s are independent, and nor-
is .7 and the probability of 1-year survival in clay soil is
mally distributed, with the following means and standard
.6, compute an approximation to P(25 X 2 Y 5) (do
deviations: m 1 5 15, s 1 5 4, m 2 5 5, s 2 5 1, m 3 5 8,
not bother with the continuity correction).