M08_LEVI5199_06_OM_C08.

QXD 2/4/10 10:56 AM Page 1

8.7 Estimation and Sample Size Determination for Finite Populations 1

8.7 Estimation and Sample Size Determination

for Finite Populations
Estimating the Mean
In Section 7.6, you used the finite population correction (fpc) factor to reduce the standard
error by a value equal to 1(N - n)>(N - 1). When developing confidence interval estimates
for population parameters, you use the finite population correction factor when samples are
selected without replacement from a finite population. If more than 5% of the population is
sampled (i.e., n>N 7 0.05), the finite population correction factor significantly reduces the
width of the confidence interval. If less than 5% of the population is sampled, the finite popu-
lation correction factor has little or no practical effect on the confidence interval width.
Equation (8.12) defines the (1 - a) * 100% confidence interval estimate for the mean.

CONFIDENCE INTERVAL FOR A MEAN (s UNKNOWN)

FOR A FINITE POPULATION
S N - n
X ; ta>2 (8.12)
1n A N - 1

To illustrate the finite population correction factor, refer to the confidence interval estimate
for the mean developed for Saxon Home Improvement Company in Section 8.2. Suppose that in
this month there are 5,000 sales invoices. Using X = $110.27, S = $28.95, N = 5,000,
n = 100, df = n - 1 = 99, and, with 95% confidence, ta>2 = 1.9842. From Equation (8.12),

28.95 5,000 - 100

110.27 ; (1.9842)
1100A 5,000 - 1
= 110.27 ; 5.744(0.99)
= 110.27 ; 5.69
$104.58 … m … $115.96

In this case, because only 2% of the population was sampled, the finite population correction
has a minimal effect on the width of the confidence interval. To examine the effect of the cor-
rection factor when the sample size is more than 5% of the population size, see Example 8.10.

EXAMPLE 8.10 In Example 8.3 on page 289, a sample of 30 insulators was selected. Suppose the company
produced a population of 300 insulators. Construct a 95% confidence interval estimate of the
Estimating the population mean.
Mean Force for
SOLUTION Using the finite population correction factor, with, X = 1,723.4, S = 89.55,
Insulators N = 300, n = 30, df = n - 1 = 29, and ta>2 = 2.0452 (for 95% confidence):

S N - n
X ; ta>2
1nA N - 1
89.55 300 - 30
= 1,723.4 ; (2.0452)
130 A 300 - 1
= 1,723.4 ; 33.44(0.9503)
= 1,723.4 ; 31.776
1,691.62 … m … 1,755.18

Because 10% of the population was sampled, the finite population correction factor has a mod-
erate effect on the confidence interval estimate.
M08_LEVI5199_06_OM_C08.QXD 2/4/10 10:56 AM Page 2

2 CHAPTER 8 Confidence Interval Estimation

Estimating the Proportion

In sampling without replacement from a finite population, the (1 - a) * 100% confidence
interval estimate of the proportion is defined in Equation (8.13).

CONFIDENCE INTERVAL ESTIMATE FOR THE PROPORTION USING

THE FINITE POPULATION CORRECTION FACTOR
p(1 - p) N - n
p ; Z (8.13)
A n AN - 1

To illustrate the use of the finite population correction factor when developing a confi-
dence interval estimate of the population proportion, consider again the estimate developed for
Saxon Home Improvement Company in Section 8.3. For these data, N = 5,000, n = 100,
p = 10>100 = 0.10, and, with 95% confidence, Za>2 = 1.96. Using Equation (8.13),
p(1 - p) N - n
p ; Z
A n AN - 1
(0.10)(0.90) 5,000 - 100
= 0.10 ; (1.96)
A 100 A 5,000 - 1
= 0.10 ; (1.96)(0.03)(0.99)
= 0.10 ; 0.0582
0.0418 … p … 0.1582

In this case, because only 2% of the population was sampled, the finite population correction
factor has virtually no effect on the confidence interval estimate.

Determining the Sample Size

Just as you used the finite population correction factor to develop confidence interval esti-
mates, you can also use it to determine sample size when sampling without replacement. For
example, in estimating the mean, the sampling error is

s N - n
e = Z
1n A N - 1
and in estimating the proportion, the sampling error is
p(1 - p) N - n
e = Z
A n AN - 1

To determine the sample size in estimating the mean or the proportion from Equations
(8.4) and (8.5),

Z 2s2 Z 2p(1 - p)
n0 = and n0 =
e2 e2

where n0 is the sample size, without considering the finite population correction factor.
Applying the finite population correction factor results in the actual sample size, n, computed
as in Equation (8.14).

SAMPLE SIZE DETERMINATION USING THE FINITE POPULATION

CORRECTION FACTOR
n0 N
n = (8.14)
n0 + (N - 1)
M08_LEVI5199_06_OM_C08.QXD 2/4/10 10:56 AM Page 3

8.7 Estimation and Sample Size Determination for Finite Populations 3

In determining the sample size for Saxon Home Improvement Company, a sample size of
97 was needed (rounded up from 96.04) for the mean and a sample of 100 (rounded up from
99.96) was needed for the proportion. Using the finite population correction factor in Equation
(8.14) for the mean, with N = 5,000, e = $5, S = $25, and Za>2 = 1.96 (for 95% confi-
dence), leads to

(96.04)(5,000)
n = = 94.24
96.04 + (5,000 - 1)

Thus, n = 95.
Using the finite population correction factor in Equation (8.14) for the proportion, with
N = 5,000, e = 0.07, p = 0.15, and Za>2 = 1.96 (for 95% confidence),

(99.96)(5,000)
n = = 98.02
99.96 + (5,000 - 1)

Thus, n = 99.
To satisfy both requirements simultaneously with one sample, the larger sample size of 99
is needed.

Problems for Section 8.7

LEARNING THE BASICS 8.93 The manager of a bank that has 1,000 depositors
8.89 If, X = 75, S = 24, n = 36, and N = 200, construct wants you to estimate the proportion of its depositors with
a 95% confidence interval estimate of the population mean, more than one account at the bank. A random sample of
m, if sampling is done without replacement. 100 depositors is selected without replacement, and 30 state
that they have more than one account at the bank.
8.90 Consider a population of 1,000, where the standard a. Construct a 90% confidence interval estimate of the pop-
deviation is assumed to be equal to 20. What sample size is ulation proportion of the bank’s depositors who have
required if sampling is done without replacement if you more than one account at the bank.
desire 95% confidence and a sampling error of ; 5? b. What sample size is needed to estimate the population
proportion to within ; 0.05 with 90% confidence?
APPLYING THE CONCEPTS c. What are your answers to (a) and (b) if the bank has 2,000
depositors?
8.91 The quality control manager at a light bulb factory
needs to estimate the mean life of a shipment of 2,000 light 8.94 An automobile dealer wants to estimate the propor-
bulbs. Assume that the standard deviation is 100 hours and tion of customers who still own the cars they purchased five
that sampling is done without replacement. years earlier. Sales records indicate that the population of
a. Construct a 95% confidence interval estimate of the pop- owners is 4,000. A random sample of 200 customers
ulation mean life of light bulbs in this shipment if a ran- selected without replacement from the automobile dealer’s
dom sample of 50 light bulbs selected from the shipment records indicate that 82 still own cars that were purchased
indicates a sample mean life of 350 hours. five years earlier.
b. Determine the sample size needed to estimate the popula- a. Construct a 95% confidence interval estimate of the pop-
tion mean life to within ; 20 hours with 95% confidence. ulation proportion of all customers who still own their
c. What are your answers to (a) and (b) if the shipment con- cars five years after they were purchased.
tains 1,000 light bulbs? b. What sample size is necessary to estimate the population
proportion to within ; 0.025 with 95% confidence?
8.92 A survey is planned to determine the mean annual
c. What are your answers to (a) and (b) if the population
family medical expenses of the 3,000 employees of a large
consists of 6,000 owners?
company. The management of the company wishes to be
95% confident that the sample mean is correct to within 8.95 The inspection division of the Lee County Weights
; $50 of the mean annual family medical expenses. A small- and Measures Department is interested in estimating the
scale study indicates that the standard deviation is approxi- mean amount of soft drink that is placed in 2-liter bottles at
mately $400. How large a sample size is necessary if sampling the local bottling plant of a large nationally known soft-
is done without replacement? drink company. The population consists of 2,000 bottles.
M08_LEVI5199_06_OM_C08.QXD 2/4/10 10:56 AM Page 4

4 CHAPTER 8 Confidence Interval Estimation

The bottling plant has informed the inspection division that 8.96 A stationery store wants to estimate the mean retail
the standard deviation for 2-liter bottles is 0.05 liter. value of the 300 greeting cards that it has in its inventory.
a. Construct a 95% confidence interval estimate of the pop- a. Construct a 95% confidence interval estimate of the pop-
ulation mean amount of soft drink per bottle if a random ulation mean value of all greeting cards that are in its
sample of 100 2-liter bottles selected without replace- inventory if a random sample of 20 greeting cards
ment indicates a sample mean of 1.99 liters. selected without replacement indicates a mean of $2.55
b. Determine the sample size necessary to estimate the and a standard deviation of $0.44.
population mean amount to within ; 0.01 liter with 95% b. What is your answer to (a) if the store has 500 greeting
confidence. cards in its inventory?
c. What are your answers to (a) and (b) if the population
consists of 1,000 bottles?