CB2200 Business Statistics

Topic 5
Confidence Interval Estimation
Tutorial Week 8

1
 Concept Review
 Confidence Interval Estimation for the Population Mean
 Using Distribution ( Known)
 Using Distribution ( Unknown)
Next week (Week 9)
 Determining Required Sample Size for Estimating Mean

2
What is a Confidence Interval?
 Provides a range of values
 Using one sample to indicate population
 Recall that if we calculate every possible samples.

 We could use one to estimate the , considering variations.

 Stated in terms of level of confidence,

 = (1 – level of confidence)
 e.g. 99% level of confidence -> = 1%
 Can never be 100% sure!

3
Confidence Interval Estimation
1. We need the to be a normal distribution
Confidence
Intervals for μ

X ~ Normal distribution X ~ other distribution

σKnown
Known Unknown Large Sample Small Sample
(n ≥ 30) (n < 30)

distribution distribution
Known Unknown
yes Z
2. Do we know
t distribution distribution
no
4
CI for ( Known)
 Conditions
 Population standard deviation () is known
 Population is normally distributed
 If population is not normal, but with a large sample , by Central
Limit Theorem
 Confidence interval estimate

Standard Error,
P(Z<
Sampling Error,
(Also called margin of error)
Critical Value

5
CI for ( Known)
 Level of Confidence
 Confidence that the interval will cover the unknown population
mean
 -value (Critical Value)
 is the value corresponding to an upper-tail probability of from
the standardized normal distribution
 Sampling Error (Margin of Error)
𝛼 /2

0 𝑍 𝛼/2 𝑍
6
CI for ( Known) Recall:
(< < )
 Derivation 2
𝜎
𝑋 𝑁 (𝜇 𝑋 , ( ))
√ 𝑛 α/2 α/2
1 −𝛼

−∨𝑍 𝛼 /2∨¿ ¿ 𝑍 𝛼/ 2∨¿ 𝑍

0
->
->
->

-> The CI for (% level of confidence is

->

7
CI for ( Known)
 Interpretation

√ If all possible samples of size 15 are taken and the corresponding



99% confidence intervals are constructed, 99% of these intervals

will cover the unknown population mean

√ We are 99% confident that the population average number of



shares traded on the Hang Seng Index is between 85351.88 and

344648.12

×  There is 99% chance that the unknown population mean will be

in between 85351.88 and 344648.12

8
Factors Affecting Interval Width
(Precision) Recall:
 Data variation Width = upper – lower
 Measured by = Xbar + E – (Xbar – E)
 width of interval =2*E

 Level of confidence
 Measured by
 -valuewidth of interval

 is a fixed parameter
 affects the location of the interval, but not the width

9
Student’s -distribution
 The variable no longer follows the standardized normal
distribution
 Mathematically, follows a distribution called Student’s -
distribution
 Also simply called -distribution
 Often denoted as
 The probability density function of -distribution is
,
where G is the gamma function and is the parameter of the
function which often called the degrees of freedom

10
Student’s -distribution
 Properties of t distribution Recall:
 For
 Mean & Standard Deviation
 Mean = 0 for >1, otherwise it is undefined
 Standard deviation = for > 2, = for , otherwise undefined
 The shape of the density function
 The theoretical range of is infinite, i.e.
 Bell shaped
 Symmetric about = 0
 Median = mode = 0
 As increases, the density curve approaches the N(0, 1) curve

11
Student’s -distribution
Cont’d

Standard Normal
(t with = ∞)

t (= 13)
-distributions have ‘fatter’
tails than the normal, i.e. t ( = 5)
more likely to get value
away from the mean

0 𝑇

12
Student’s -distribution
Cont’d

Find the here

Find degree of
freedom here The value within the table gives the t-value
corresponding to a particular degrees of freedom
and upper-tail area
At 7 degrees of freedom,
13
 CI for ( Unknown)
 Conditions
 Population standard deviation () is unknown
 Population is normally distributed
 If population is not normal, but with a large sample , by Central
Limit Theorem
 Confidence interval estimate

Critical Value Standard Error

 With the use of -distribution with degrees
Sampling Error of freedom
in this context

14
Factors Affecting Interval Width
(Precision)
 Sample Size
 width of interval
 width of interval
 Level of confidence
 Measured by
 -valuewidth of interval
 is a random variable
 affects the location of the interval, but not the width

15
Determining Sample Size
 Statisticians have control over the sampling error () by
choosing appropriate sample sizes that are large enough
to make the results appear credible
 Sampling error measures how far off the estimation results are
likely to be from the result that they would have gotten if the
entire population are surveyed instead of merely a sample
 never use t value!
 The required sample size can be
found to reach desired sampling
error with a specified level of
confidence 𝑋
𝜎 𝑋 + 𝑍 𝛼/ 2 𝜎
𝑋 − 𝑍𝛼 / 2 √𝑛
√𝑛
= Sampling Error
16
Determining Sample Size
Cont’d

 What sample size is needed to be confidence of being

correct to within ?
 Assume is known

Round up to nearest integer

17
Summary

Week7

Week8

18