Remark 1.

For some special condence levels, we have:

For = 0.90, Z0.05 1.645.
For = 0.95, Z0.025 1.96.
For = 0.99, Z0.005 2.58.
Remark 2. The meaning of the C.I.: The C.I. obtained with condence level
has a probability of to cover the unknown population .
Remark 3. We should NOT say that has a probability of to be contained in
the C.I. obtained: because is a xed unknown constant and is NOT a random
variable but the end points of the C.I. are random variables.
the population variance 2 and the conRemark 4. Given the sample mean X,
dence level , the larger the sample size n is, the shorter the C.I. will be and
the length will tend to zero as n tends to innity.
Remark 5. Given the sample mean, the population variance and the sample size,
the larger the condence level is, the longer the C.I. will be.
1

Example 1. The average demand of certain new product is (unknown) in the

network of department stores. A survey of 40 random samples yields a mean of
600 per day. From the past experience of new products, the variance 2 is known
to be 400. Construct a 90% and 95% condence intervals for the mean .
= 600, = 20 and n = 40. Then:
From the assumptions, we have X
For = 0.90, we rst obtain Z0.05 = 1.645 and then give the C.I. as follows:
]
[
20
20
= [594.8, 605.2].
600 1.645 , 600 + 1.645
40
40
For = 0.95, we rst obtain Z0.025 = 1.96 and then give the C.I. as follows:
]
[
20
20

, 600 + 1.96
= [593.8, 606.2].
600 1.96
40
40
Exercise 1. In a public exam, a sample of 100 scores yields an average of 57.4.
From the past experience, the standard deviation is around 10. Construct a 90%
and 95% condence intervals for the average score .

16.1

Case of Unknown Variance 2

We know how to construct a C.I. for the population mean at a given condence
level when the population variance 2 is known. The key idea is that by the
central limit theorem, for large n,

X

Z=
/ n
follows the standard normal distribution approximately.
If the population variance is unknown, with n random samples, {X1, X2, , Xn},
we can dene a new random variable:

X

T =
(16.1)
S/ n
where

1
1
2
2.

Xi and S =
(Xi X)
X=
n i=1
n 1 i=1
n

Recall that S 2 is called the sample variance and it is a point estimate for the
population variance 2.
T in (16.1) follows Students t-distribution with degree of freedom (n
1). One can construct C.I.s for dierent condent levels by using Students tdistribution.
This probability distribution will NOT be covered in this course. However, it can be
shown that for sample size n 30, Students t-distribution with degree of freedom
n can be well approximated by the standard normal distribution N (0, 1).
So, as often done in practice, when the sample size is greater than or equal to
30, to construct the condence intervals for the case with unknown population
variance, you may still adopt the normal distribution to replace Students
t-distribution.

Example 2. We repeat the previous example. The average demand of certain new
product is (unknown) in the network of department stores. A survey of 40 samples yields a mean of 600 per day and sample standard deviation is 18. Construct
a 90% and 95% condence intervals for the mean .
= 600, S = 18 and n = 40. Then:
From the assumptions, we have X
For = 0.90, we rst obtain Z0.05 = 1.645 and then give the C.I. as follows:
[
]
18
18
600 1.645 , 600 + 1.645
= [595.3, 604.7].
40
40
For = 0.95, we rst obtain Z0.025 = 1.96 and then give the C.I. as follows:
[
]
18
18
600 1.96 , 600 + 1.96
= [594.2, 605.6].
40
40
Exercise 2. In a public exam, a sample of 100 scores yields an average of 57.4
and the sample variance is 8. Construct 90% and 95% condence intervals for the
mean .

17

Inference for Proportion

Suppose n samples are obtained from a Bernoulli population (positive and

negative) and the proportion of positives is p (unknown).
Let Y be the number of positives in the sample of size n, then Y follows the binomial distribution Binomial(n, p) with mean np and variance np(1 p).

By the central limit theorem, for large n, (Y np)/ np(1 p) approximately

follows the standard normal distribution N (0, 1) (and therefore, Y approximately
follows the normal distribution N (np, np(1 p))).
Since p is unknown, we further approximate p by p , Y /n in the variance by
np(1 p). Then we have
Y np
Y np
p p

=
,

p (1p )
np(1 p)
np (1 p )
n

which, again, approximately follows the standard normal distribution N (0, 1).
6

To construct a C.I. for p, we consider








 p p 

P (
Z/2 = ,

 p(1p) 


n
which gives the following C.I.:
[
]

p(1 p)
p(1 p)

p Z/2
, p + Z/2
.
n
n
Example 3. To study the proportion of smokers in a population, a survey has
been conducted and a sample of 1000 has been obtained. Among them 170 are
smokers. Construct a 90% C.I. for the smoker proportion.
Now we have n = 1000, Y = 170 and therefore p = 0.17 and Z/2 = 1.654. Thus
the 90% C.I. is
[
]

0.17 0.83
0.17 0.83
0.17 1.645
, 0.17 + 1.645
or [0.1504, 0.1895].
1000
1000
7

A Summary

1. Point estimate of population mean and variance.

2. Interval estimate of population mean.
3. Sample variance.
4. Condence interval for population mean with known variance.
5. Condence interval for population mean with unknown variance.
6. Inference for proportion.