Chapter 8 – Confidence Intervals

In this chapter we are entering the part of statistics that is inferential statistics.
The purpose of collecting data on a sample is not only to describe the sample but
to be able to use this data and infer (generalize) to the population represented by
this sample. Inferential Statistics is the collection of methods that help us make
decisions on the population based on sample data

In this chapter we are going to learn how to

 Construct a Confidence Interval for the population mean µ

 When the Population standard deviation ϭ is known
 When the Population standard deviation ϭ is unknown
 Construct a Confidence Interval for the population proportion P
 Determine the required Sample size

Estimation : is a procedure by which a numerical value or values are assigned to a

population based on sample values

In Statistics µ is called the true population mean and we use x the sample mean to
estimate it, and P is called the population true proportion and we use ^pthe
sample proportion to estimate it.

Point Estimate: If we pick a random sample and calculate its mean x then this will
be the point estimate of the population value µ.

Point Estimate of a population parameter = Value of the corresponding sample

statistic

But remember from Chapter 7 when we picked all possible samples each sample
yielded a different x . So the actual value depends on which sample was picked.
the point estimate then assigns a value for µ which is almost always different
than the true population mean.
Confidence Interval Estimate of a population mean
So now we are going to find an estimate for the population value, this value is a
range based on data from a sample with a level of confidence. This interval is
constructed around the point estimate, and a Level of Confidence that this
interval will catch the parameter is made.

Before we can construct a Confidence Level there are certain conditions that have
to be met
1. The sample must be randomly selected
2. The sampling distribution of x should follow a normal distribution
Remember from Chapter 7 for the sampling distribution of x to follow
a normal distribution
 Either the variable of interest x has a normal distribution in the
population
 Or the sample size ≥ 30

Confidence Interval for µ is :

x ± Z ϭx
ϭ
Where ϭ x = n
√

Z you find from the Z table depending on the confidence level

To find the critical value, or Za/2: Here, the
confidence level is 95%. Convert the percentage to
a decimal, 0.95, then subtract is from 1 and the
difference divide by 2 gives us the value 0.025
Then, check out the z table to find the
corresponding value that goes with 0.025 You'll
see that the value is 1.96, at the intersection of
row 1.9 and the column of .06.
 Z1 =-1.96
Z2 = 1.96
We take the value of Z without the sign the sign is already in the
formula.

The most commonly used confidence levels are 90 percent, 95

percent and 99 percent.

The Value Z ϭ x is called the Margin of Error and is denoted by E or

ME
The Margin of error is the maximum likely difference between the
observed sample mean and the true value of the mean that is the
maximum error of the estimate

So we can write the Confidence interval as

x± Margin of Error

The two values we get are called

Upper limit : x + Z ϭ x or x + ME
Lower limit : x - Z ϭ x or x – ME

The width of the Confidence interval is the difference between the upper Limit
and the lower limit
W = Upper Limit – Lower Limit = 2 Z ϭ x = 2 ME

There are 3 factors that could affect the width of the confidence Interval

 The Confidence level : The larger the confidence level is , the larger is the
absolute value of Z
( for example 95% z= 1.96 ; 99% z= 2.575 ), so as a result the larger is the
Margin of error ( keeping everything else the same ) and the wider is the
confidence interval
ϭ
 The larger the sample size n the smaller is ϭ x (remember ϭ x =¿
√ n ) as a
result the narrower is the confidence interval
 The third factor is the population standard deviation but this is a property
of the population we are studying and we have no control on ( but if we are
comparing 2 populations with everything the same the larger the
population standard deviation ϭ the larger the width of the confidence
interval )

The width of a confidence interval may be decreased by lowering the confidence

level or by increasing the sample size. Since lowering the confidence level results
in a less reliable estimate, it is preferable to increase the sample size.

Example : Exercise 8.20

n = 1500

sample mean = $287,530

Population standard deviation =$ 53,870

Construct a 99% confidence Interval

Since n is very large the sampling distribution of x follows a normal distribution

Confidence level is 99% so the area is (1-0.99)/2) = 0.005 so the Z is 2.575

ϭ 53,870
Margin of error = Z = 2.575
√n √1500
287,350 ± 3,581.61 = $283,768.39 to $290,931.61 =

$283,768.39< µ < $290,931.61

Interpretation of the confidence level:

If we take all possible samples of n=1500 and construct a 99% confidence interval
for µ from each sample, we can expect about 99% of these confidence intervals
will contain µ and 1% will not.

(For more Details on interpretation refer to the handout – interpretation of

Confidence Interval)

Z values for the most common confidence levels

a. 90% confidence level z = 1.645

b. 95% confidence level z = 1.96
c. 96% confidence level z = 2.05
d. 97% confidence level z = 2.17
e. 98% confidence level z = 2.33
f. 99% confidence level z = 2.575

Sample Size determination

How do we decide how large a sample we should pick

From what we have already covered we learned that the
ϭ
margin of Error = Z n
√
So if we decide on our margin of error and we want to determine the sample
size which will yield this margin of error
ϭ
Margin of error = Z
√n
Zϭ
n=( E )2
Based on a confidence level we chose and remember that ϭ is known to us we
can determine n
Example : Determine the Sample size for
 E=1.45 ϭ= 5.82 confidence level =95%
Zϭ
n=( E )2
95% confidence level : we need to find the Z with the area to the left = (1-
0.95)/2=0.025 this Z is 1.96
1.96 x 5.82 2
n=( 1.45
) = 61.89= 62

** Rule of rounding in sample size determination : We always rund upward

irrespective of what the decimal value is that is even if n= 61.29 we would
round it to 62

Case II : The population Standard Deviation ϭ is unknown

Now we are going to see how to construct a confidence interval in the case where
the population standard deviation is not known

The conditions that the sample should be a random sample and that the sampling
distribution of x follows a normal distribution should be met.

When the standard deviation of the population we are studying is unknown we

are going to replace Population Standard deviation (ϭ) by the sample standard
deviation (s) and instead of using Z coefficient we are going to replace it with
the t coefficient

The t-distribution is used when the sampling distribution of x is normally

distributed, but the population standard deviation is unknown.

t-Distribution vs. normal distribution

 Like the normal distribution, the t-distribution is bell shaped, that is
unimodal and symmetrical.
 Both, the standard normal distribution (or z-distribution), the t-distribution
have a mean of zero. So like the Z values we can have negative and positive
values.
 The normal distribution assumes that the population standard deviation is
known. The t-distribution does not make this assumption. The standard
deviation of the Z distribution is 1 while that of the t – distribution is ≥1
As such the t- distribution is more spread out (refer to the figure below)

 The t-distribution is defined by the degrees of freedom. These are related to

the sample size. (remember we discussed the degrees of freedom at the
beginning of the course when we were studying how to calculate the
sample standard deviation)
 As the sample size increases, the t-distribution becomes more similar to a
normal distribution.
To read the critical value from the t distribution we are going to use the APP

V: the degrees of freedom

In this case the Degrees of freedom are: n -1 since we are dealing with one
sample. ( in the following chapters we will see different ways of calculation
degree of freedom)

So if confidence level is 95% and the sample size is 10 then the t coefficient we

In the rectangle next to 2P(X>|x |) we insert 0.05 ( which is 1 – confidence level,

it is the total areas in both tails)

degrees of freedom: V = n-1 =10-1=9

t=2.262.

So the confidence level now becomes

Confidence Interval for µ is :

x ± ts x
s
Where s x =
√n
And the margin of error is now ts x

Example

8.43 A random sample of 20 acres gave a mean yield of wheat equal to 41.2
bushels per acre with a standard deviation of 3 bushels. Assuming that the yield
per acre is normally distributed, construct a 90% confidence level for the
population mean

Solution :

n = 20, x = 41.2 bushels per acre,

s = 3 bushels per acre, We know that this is a sample std dev and not population
from the given because they say a sample gave a mean and a std dev

The variable is given normally distributed as a result the sampling distribution of x

is normally distributed.
s 3
sx = = = 0.67
√ n √20
90 % confidence level so the area is (1-0.9)/=0.1

Df =n-1 = 20-1 =19

t= 1.729

The 90% confidence interval for μ is x ± ts x :

41.2 ± 1.729(.67)= 41.2 ± 1.15 = 40.05 to 42.35

Confidence Interval For Population Proportion

Remember for Chapter 7 regarding the sampling distribution of ^p :

Mean of the sampling distribution of ^p is : μ p^ = p

√
Standard Deviation of the sampling distribution of ^p is : σ ^p = pq
n
Shape of the sampling distribution of ^pis normally distributed if np≥ 5 and

nq is ≥5

But when estimating the population proportion we do not know p we are trying
to estimate it going to replace p by ^pand so σ = pq
^p
√ n
Becomes s ^p =
√ ^p q^
n
And the confidence interval for the population proportion p is :

^p ± Z s ^p = ^p ± Z
√ ^p q^
n

Where Z is obtained from the standard normal distribution table

^^
√
And Z p q is the margin of error which is the maximum error of the estimate
n

Example. Exercise 8.69

In a random sample of 50 homeowners selected from a large city , 19 said they

have a problem with excessive noise from their neighbours .

Solution :

a. n = 50, ^p= 19/50 = .38, q^ = 1 - ^p= 1 - .38 = .62,

n ^p = 50 x 0.38 = 19 and n q^ b = 50 x 0.62 = 31
Both are greater than 5 so the sampling distribution of ^pis normally
distributed

s ^p =
√ √
^p q^
n
= 0.38 x .0 .62
50
= 0.069

99% confidence Level so the area is (1-0.99)/2 = 0.005

Z = 2.575
The Confidence interval is : ^p ± Z s ^p = 0.38 ± ( 2.575)(0.069)= 0.38 ± 0.18
The 99% confidence interval is 0.20 to 0.56

b. The width of the confidence interval constructed in part a may be reduced by:
1. Lowering the confidence level
2. Increasing the sample size
The second alternative is better because lowering the confidence level results in
a less reliable estimate of p.
Determining the Sample Size for the estimation of proportion
Just as we did previously with the mean we can find what is the sample size
needed for estimating P with a certain Margin of Error and confidence Level

√^^
Margin of Error = Z p q
n
Z is found from the standard normal table depending on the confidence level
2
Z p^ q^
n= E
2

In the case where ^p is not known we make the most conservative estimate of the
sample size n by using ^p = 0.5 and q^ = 0.5. Foe a given Margin of error these
values give us the largest sample size since their product 0.5x0.5= 0.25 is greater
than any other pair

Example
Determine the sample size for the estimation of the population proportion for the
following
E = 0.04 ^p = 0.78 confidence level 95%

2
Z p^ q^
n= 2
E
Z from the standard normal distribution table for the area 0.025 is 1.96

2
1.96 (0.78)(0.22)
n= =412.01= 413
0.042