Chapter 8

Confidence Interval
Estimation

David Chow
Oct 2021

1
 Learning Objectives

 To construct and interpret confidence interval

estimates for the mean and the proportion
 To determine the necessary sample size for a
confidence interval
 Section 8.5: Applications in auditing (NOT covered)

2
 Basic Concepts
 A point estimate is a single number
 A confidence interval is an interval estimate -- indicates variability

Eg: Interval Estimate

 According to a survey, the mean wage of
tutoring is between $120 to $180 per hour
μ = $150 ± 30

3
 Basic Concepts
 The general formula for all confidence intervals (CI) is:

Interval Estimate (CI) = Point Estimate ± Margin of Error

 Two main
applications:

4
 Basic Concepts

 Let’s begin with estimating 

 Easiest case: estimating  with a known 
 A more realistic case ( unknown) follows

 Concepts versus Computation

 As always, concepts can be a bit abstract, but computations are
straightforward
 Let’s master the computations first, then go back to the rationale
and interpretation

5
Estimating μ
(σ Known)

6
 Estimating μ
 Assume population standard deviation σ is known
 Also assume n is large enough (n > 30)
or a normally distributed population
 These assumptions ensure that ____

 A two-tailed confidence interval estimate:

σ
XZ
n

7
 Critical Values of Z
 Z, also written as Zα/2, is the standardized normal distribution
critical value for a probability of ____ in each tail
 Z is associated with a confidence level of (1-α) %

 Consider a 95% confidence interval:

Critical Z-value(s)

8
 Critical Values of Z

 With a confidence level of 95%, Zα/2 = ____

 More Examples

 Z0.05 corresponds to a confidence level of ___%,

 Z0.05 = ___

 Z0.005 corresponds to a confidence level of ___%,

 Z0.005 = ___

9
 Eg: Length of A4 Paper
 A paper producer wants to check if the output
has the correct mean length of 11 inches
 Find the 95% CI for the mean paper length

• Sample info
• Sample size n = 100
• Sample mean = 10.998 in
•  is known to be 0.02

10
 Eg: Length of A4 Paper
 The 95% confidence interval is given by:

 =x  Z/2  x

 Step 1: Find Z0.025 =  1.96

 Step 2: Z/2  x = 1.96 (0.02)/10 = 0.00392

 The required CI is:

10.99408 < µ < 11.00192
  = 10.998  0.00392 inches, or

Conclusion?

11
Eg: Mean Resistance

 A sample of 11 circuits has

a mean resistance of 2.20
σ
ohms X  ZANSWER
0.025
n
 σ = 0.35 ohms from past
 2.20  .2068
testing
(1.9932 , 2.4068)
 Determine a 95% CI for
the true mean resistance
 We are 95% confident that the true mean is
of the population
between 1.9932 and 2.4068 ohms
 What assumption is  The population is assumed to be normally
needed from the distributed as n=11
population?
 Remember the “Central Limit Theorem”?

12
Eg: Mean Resistance (Continued)

 Is it correct to use the Z-distribution?

 x is called standard error. Why is it so?

Recall
σ
 XZ
n

Margin of error = Z ∙ (standard error)

13
 Choosing Confidence Level
 Confidence level is our choice
 Common picks: 90%, 95% or 99%

 Eg: Length of A4 paper

 Recall 95% CI: µ = 10.998 ± 0.00392 inches, or 10.99408 < µ < 11.00192
 Now find the 99% confidence interval
 What is the effect of raising the confidence level?
a) We have more confidence (whatever that means) about the answer
b) The confidence interval is wider
c) The margin of error is bigger
d) All of the above

14
 Choosing Confidence Level
 Confidence level is our choice
 How to choose?

 A bigger confidence level raises the confidence

 But a wider interval estimate also means ____ precision
 95% is the most common choice
 It provides a good balance
between precision and confidence

15
α, Confidence Intervals &
Sampling Distribution

/2 1  /2

x
μx  μ

x1
x2  (1-) x 100% of
intervals constructed
contain μ;
 () x 100% do not

Confidence Intervals
16
 Interpreting Confidence Intervals

Repeated Sampling
 If we select many different samples of size n from a population,
 and construct a 95% confidence interval for each sample

 Then 95% of those interval estimates would actually contain (or

cover) the true value of μ

Naturally, 5% of the intervals would

not contain the population mean

Questions

17
 Estimating μ
(σ Unknown)

In the previous section

• σ is known,
• but µ is unknown

Question

18
 Estimating μ (Unknown σ)
 Naturally, we use sample SD (S) to replace σ
 Extra uncertainty introduced by S
 S varies from sample to sample
 The random variable t is defined as

 It follows the t distribution with n-1 degrees of freedom

 Hence

19
 t Distribution
 The t distribution is a
family of curves
 t distributions are:
 Bell-shaped and symmetric
 Characterized by
degrees of freedom
 In this chapter,
 A specific t distribution is
df = n-1
associated with a particular df

20
 t Distribution and d.f.
•• tt distributions
distributions look
look like
like the
the Z
Z curve
curve
•• tt distributions
distributions are
are generally
generally flatter
flatter than
than Z
Z
•• The
The larger
larger the
the sample
sample size,
size, the
the more
more the
the tt distribution
distribution looks
looks like
like ZZ
•• The
The difference
difference between
between tt distribution
distribution and
and ZZ disappears
disappears when
when ____
____

21
 Critical Values of t
Suppose n = 21, and  = 0.10
Then df = ____
upper-tail area = ____

d.f. = 20

/2 = 0.05

The body of the table contains t values

 In a Z table, the body contains ____
0 1.724 t

22
 Estimating μ (Unknown σ)
 Assumptions: With an unknown σ, we need
(1) a normal population, or  For a highly-skewed population, or when
(2) n ≥ 30 outliers exist, n shall be 50 or more
 If the population is fairly symmetric,
n = 15 can be enough

 CI estimate:   X  t S
n -1
n
 where t, also written as tα/2,n-1, is the critical t-value

n-1 indicates degrees of freedom,

and α/2 is the upper-tail area

23
 Eg: Mean Age of Retirement
 A random sample of 25 retirees has mean age = 50 and std = 8
 Find the 95% confidence interval for 
 Assume ____

 From t-table, t0.025, 24 = ± 2.0639

 The confidence interval is

S 8
X  t/2, n -1  50  (2.0639)
n 25

(46.698 , 53.302)

24
 Critical t Values Again
 Critical t values depend on two elements:
 The confidence level (1- ), and df

What is this
Degrees Area in Upper Tail “2.009”?

of Freedom .20 .10 .05 .025 .01 .005

. . . . . . .
50 .849 1.299 1.676 2.009 2.403 2.678
60 .848 1.296 1.671 2.000 2.390 2.660
80 .846 1.292 1.664 1.990 2.374 2.639
100 .845 1.290 1.660 1.984 2.364 2.626
.842 1.282 1.645 1.960 2.326 2.576

Look familiar? What are these circled values?

25
Eg: Heating Oil Consumption
 A random sample of 35 households has mean consumption of heating
oil = 1122.75 gallons, and S = 295.72 gallons
 Find the 95% confidence interval for 
ANSWER
 ANSWER
 Critical values are  t0.025, 34 =  2.0322.
  = 1122.75  101.58 gallons.
 Based on the sample evidence, we are 95% confident that the interval
1122.75  101.58 gallons covers the population mean.
NOTE: Z or t?
 Check:
1. σ given or not?
2. Distribution of ?
 If n  30, it is commonly acceptable to use Z (instead of t) as an approximation
 But t-values give us a more precise answer 26
 Estimating
Population Proportion

27
 Confidence Intervals for π

 Recall that the distribution of the sample proportion is

approximately normal if nπ ≥ 5 and n(1-π) ≥ 5

 Standard deviation  (1   )
is given by σp 
n

 As π is unknown, we will
p(1 p)
estimate with sample data:
n

28
 Confidence Intervals for the
Population Proportion π
The confidence interval for the population proportion is given by:

p(1  p)
  pZ
n
where
 Z = critical Z-value given the level of confidence
 p = sample proportion, n = sample size

Like μ, this interval estimate for π is based on a point estimate (p),

plus an allowance for uncertainty from sampling

29
 Example: Vaccinations
1. A random sample of 100 people shows that 25 of them are vaccinated.
Form a 95% confidence interval for the population proportion
2. Compute the 95% confidence interval if n=1000

 Interpretation
 95% of intervals formed from
samples of size 100 in this manner
will cover the true proportion

 Sample Size
 If n=1000, margin of error = ____
 For n=100, π = 25% ± 8.5%
 For n=1000, π = 25% ± 2.7%

30
Sample Size
Determination

31
 Sample Size Determination
 Recall that sample size (n) affects the margin of error (e)
σ
 e is also called sampling error: e  Z
n

 If e is set before conducting a survey, this equation helps you

determine the sample size for a pre-set value of e
 e can be regarded as the acceptable level of error:

Z2 σ 2
n
e2

32
Sample Size Determination

If  = 45, what sample size is needed to estimate

the mean within ± 5 with 90% confidence?

Z 2 σ 2 (1.645) 2 (45)2
n 2
 2
 219.19
e 5

Round up to the next integer to get

the required sample size n = 220

33
 Eg: A4 Paper Again
 Recall the A4 paper example where  = 0.02 and n = 100
 The 95% interval estimate is  = 10.998  0.00392 inches
 Suppose the manufacturer wants to limit the error to 0.003 by
choosing a larger sample. What is n?
ANSWER
 ANSWER

 The required sample size is n = 171.

Z 2 σ 2 (1.96) 2 (0.02) 2
n 2
 2
 170.7
e 0.003

34
 Sample Size Determination

To find the required sample size for the proportion, you must know:
 The critical value Z (from a confidence level of 1-α)
 The acceptable sampling error (e), and
 The true proportion π
 If π is unknown, use the sample value p, or set π = 0.50

 (1   ) Solve for Z 2  (1   )
eZ n to get n
n e2

35
Eg: Quality Control
 Out of a population of 1,000 light bulbs, we randomly selected
100 of which 30 were defective. What sample size is needed
to be within ± 0.05 with 90% confidence?
(a) As the population proportion is unknown, use the sample value
(b) Now, set π = 0.50 and compare the result with (a)

ANSWER
(b) The required sample size
(a) Z p 1  p 
2
1.645  0.3 0.7 
2
increases to 271
n 
Error 2 0.052
 227.3  228 NOTE: The product π (1- π) ranges
from 0 to 0.25. By assuming a value
of 0.25, we are in fact playing safe by
sampling more than necessary.
36
More on the
t Distribution

37
 t Distribution

The t distribution is a family of probability

distributions. It is bell-shaped, symmetric, &
flatter than the Z distribution

A specific t distribution depends on a parameter

known as the degrees of freedom (d.f.)

Degrees of freedom refer to the number of

independent pieces of information that go into the
computation of s

38
 Degrees of Freedom
 The critical value of t is characterized by two elements:
 The confidence level (1- ), and
 The degrees of freedom (df)

 What is d.f.?
 It is the number of observations that are free to vary
after sample mean has been calculated
 In this section, df = n-1

39
 Degrees of Freedom
Eg: Suppose the mean of 3 numbers is 8.0
 Let X1 = 7, X2 = 8
 What is X3?

 Given a mean value of 8.0, X3 must be 9

 In other words, X3 is not free to vary

 In this example, d.f. = n-1 = 2

d.f.
You= 2are
…“free” to choose 2 values (X1 and X2),
but the
What doesthird is set for a given mean
it mean?

40
Degrees of Freedom

t Z as n increases

41
Review Questions
 A random sample of 100 from a population is selected, and is 600. σ is
known to be 50. To find the 95% CI, we use the ___ table
A) Z, because sigma is known
B) Z, because n is greater than 30
C) t
D) Both (A) and (B)

 As the number of degrees of freedom for a t distribution ____, the difference

between the t and the Z distributions becomes smaller
 In a random sample of 144 observations, sample proportion p is 0.60. The
95% confidence interval for π is
a. 0.52 to 0.68
b. 0.144 to 0.200
c. 0.60 to 0.70
d. 0.57 to 0.63
Eg: Bottled Water

 A consumer organization wants to estimate the amount of water

contained in 1-gallon bottles
 σ is given at 0.02 gallon
 A random sample of 50 bottles is selected, and is 0.995 gallon

a. Construct a 95% confidence interval for the population mean

amount of water
b. From the result of (a), do you think the consumer organization
has a right to complain to the water bottling company?
c. Do you have to assume that the population amount of water per
bottle is normally distributed here?
 ANSWER
a. The 95% CI is given by:

OR: 0.9895 ≤ µ ≤ 1.0005

b. Since the specified value of 1.0 is covered by the interval,

there is no reason to complain

c. No. Since σ is known and n > 30, we may assume that the
sampling distribution of ____ is approximately normal