Chapter 8

Confidence Intervals
Learning Objectives
• Distinguish between point estimates and interval estimates
• Understand the basic structure of a confidence interval
• Interpret confidence intervals
• Understand how confidence level and sample size affect the margin
of error
• Compute a confidence interval for a population mean
• Use Z or T critical values appropriately in confidence interval formulas
• Compute a confidence interval for a population proportion
• Determining the sample size for a desired margin of error
Point Estimates vs. Interval
Estimates
• A point estimate for a population parameter is a single statistic
value, like a sample mean or sample proportion.
• Example: A survey of 400 recent college graduates with social
science degrees determines that their average starting salary is
$46,800. Based on this sample we estimate that the actual mean
starting salary of all recent college graduates with social science
degrees is $46,800.
• An interval estimate for a population parameter is a possible range
of values based on the value of a point estimate and a measure of its
variability.
Interval Estimates
• The measure of variability of a point estimate is called the margin of
error (MOE).

• The length of the interval is twice the MOE, i.e the MOE is half the
length of the interval
• The point estimate is the midpoint of the interval
Example
• Based on the results of a random sample, we estimated the
average weights of male college students to be between
158 lbs and 194 lbs.
• What is the point estimate for the mean weight of male
college students?
Find the midpoint: (158 + 194) / 2 = 176 lbs

• What is the margin of error for this estimate?

Find the half the length: (194 – 158) /2 = 18 lbs
Confidence Statements
A confidence statement states:
1.the level of confidence
2.the parameter (mean or proportion) being estimated (refers to
population)
3.the interval bounds (or an estimate and its margin of error ( ± "value"))

Example: We are 90% confident that the percent change in new

residential housing starts for the U.S. in April 2020 is between 19.2% and
41.2% below the previous month's estimate.
Level of Confidence
• A level of confidence is stated
as a percent and represents the
percent of similar intervals that
we would expect to contain the
true value of the population
parameter if we created
intervals from many samples.
Level of Confidence
• The level of confidence is used
to determine the critical value
needed to compute the margin
of error.
• Example: For a 90% confidence
interval, we need to find the
values on our probability
distribution that mark the
boundaries of the middle 90%.
Computing the margin of error

In some situations, we will use a critical value from a normal distribution,

called a Z critical value

In other situations, we will use a critical value from a t-distribution.

Effect of confidence on the MOE
Increasing the confidence level, increases the critical value
and increases the margin of error, expanding the interval
width. Note that the interval has the same center.

Z = 2.576

Z = 1.96
The Standard Error
• In chapter 7, you learned about the standard error for two statistics,
the sample mean and the sample proportion
and

In this chapter we often don’t know the values of σ or p, so we use

sample estimates of these values (s and .
The impact of sample size on MOE
• Sample size is included in the calculation of the standard error,
specifically, in the denominator. This means that as the sample size
increases, the standard error will decrease.
• Since the MOE = (critical value) x (standard error) if we decrease the
standard error, the MOE will also decrease.
Increasing the sample size, reduces the margin of error,
narrowing the interval width. Note that the interval has the
same center.
MOE for a Population Proportion
MOE for a population mean

To estimate a confidence interval for a population

mean the margin of error is calculated as:
Example: Finding t-critical values
A simple random sample of size 10 is drawn from a normal population. Find the
critical value for a 95% confidence interval.
Solution:
The sample size is n = 10, so the number of degrees of freedom is n – 1 = 9. We
consult our t-distribution table, looking in the row corresponding to 9 degrees of
freedom, and in the column with confidence level 95%. The critical value is =
2.262.
Confidence interval for Population
Mean
Example 1
• An SRS of 30 fifth grade students was chosen to participate in a
program to improve math skills. The students were tested before and
after the program. The sample mean increase in the exam score was
12.2 points with a standard deviation of 4.7 points. Determine a 95%
confidence interval for the mean increase in score as a result of this
program.
Finding the critical value
• Determine the critical value for Example 1: n =30 95% CI for a mean
Example 1 (Solution)
• An SRS of 30 fifth grade students was chosen to participate in a program to
improve math skills. The students were tested before and after the program.
The sample mean increase in the exam score was 12.2 points with a standard
deviation of 4.7 points. Determine a 95% confidence interval for the mean
increase in score as a result of this program.
• Point estimate of mean = 12.2
𝑠
• Critical value: t= 2.045 𝑥±𝑡
√𝑛
• Standard deviation: s = 4.7 (from the sample)
• Sample size: n = 30

= (12.2 – 1.75, 12.2 + 1.75) = ( 10.45, 13.95)

Example 2
• A survey of 300 American adults, found that 165 do not have a will
specifying the handling of their estate. Determine a 90% confidence
interval for the proportion of American adults who do not have a will.
Finding the critical value
• Determine the critical value for Example 2: n =300 90% CI for
proportion
Example 2 (Solution)
• A survey of 300 American adults, found that 165 do not have a will
specifying the handling of their estate. Determine a 90% confidence
interval for the proportion of American adults who do not have a will.

√
• Point estimate of proportion: 𝑝^ (1− 𝑝^ )
• Critical value: z = 1.645 (90% confidence) 𝑝^ ± 𝑧
𝑛
• Sample size: n = 300
Estimating Sample Size
Example 1
• In a 2018 survey, 33% of U.S. adult drivers stated that traffic was
getting worse in their community. If we wish to repeat this survey
and determine a 90% confidence interval for the proportion who
think that traffic is getting worse in their community, how large a
sample would be needed so that our interval will have a margin of
error of 8%?
Example 2
• A water utility wants to take a sample of residences in a city in order
to estimate the proportion who have reduced their water
consumption during the past year. How large a sample must be taken
so that a 95% confidence interval will estimate this proportion within
5%?
Example 3
• We wish to estimate a 99% confidence interval for the mean time
spent in bed asleep by young adults within 0.25 hours. How large a
sample would be needed, assuming the standard deviation of time
spent in bed is 1.11 hours?
Example 1: Solution
• In a 2018 survey, 33% of U.S. adult drivers stated that traffic was
getting worse in their community. If we wish to repeat this survey
and determine a 90% confidence interval for the proportion who
think that traffic is getting worse in their community, how large a
sample would be needed so that our interval will have a margin of
error of 8%?
• Prior estimate of p = .33
• MOE = 8% or 0.08 So we need to sample 94 drivers.
• 90% CI means z = 1.645
Example 2 (solution)
• A water utility wants to take a sample of residences in a city in order
to estimate the proportion who have reduced their water
consumption during the past year. How large a sample must be taken
so that a 95% confidence interval will estimate this proportion within
5%?
• Prior estimate of p = .50 (use when no specific estimate is given)
• MOE = 5% or 0.05 384.16
• 95% CI means z = 1.96 So we need to sample 385 residences.
Example 3 (Solution):
• We wish to estimate a 99% confidence interval for the mean time
spent in bed asleep by young adults within 0.25 hours. How large a
sample would be needed, assuming the standard deviation of time
spent in bed is 1.11 hours?
• 99% means z = 2.576
• Margin of error= m = 0.25
• Standard deviation, σ = 1.11 So we need to sample 131 young adults.