ECOM9009

Quantitative Methods
for Business

Dr. Wasana Karunarathne

Department of Economics
Faculty of Business and Economics
Introduction to Statistical Inference
 Learning Outcomes

LO1 Explain the importance of statistical inference.

LO2 Explain how, when and why statistical inference is used.

LO3 Understand the probability link between sample statistics and population
parameters.

LO4 Understand the central limit theorem and the properties of the sampling
distribution of the sample mean.

LO5 Understand the properties of the sampling distribution of the sample

proportion.
3
 What we cover today!

So, we know what we do!

• Simple Random Sampling
• Sampling Distribution for Sample mean
• Sampling Distribution for Sample Proportion
• Central Limit Theorem

4
 Revise: Population Vs Sample

• Population is the full set of all items that Examples:

are of interest in drawing inferences. 1. Students enrolled in QMB in
2025, Semester 2.
• Characteristics of a population are called
2. Registered voters for the 2025
parameters. E.g.:
federal election in Australia.
• Population Mean ( )
3. Customers who have purchased
• Population Proportion ( ) from Amazon, an online retail
store, on a given day.
• Population Standard Deviation ( )

5
 Revise: Population Vs Sample

• Sample is a subset of data drawn from a Examples:

population. 1. Randomly selected 20 QMB
students from 2025 Semester 2.
• Characteristics of a sample are called
2. A representative sample of
Statistics. E.g.:
2,097 Australian electors.
• Sample Mean ( ) 3. A sample of 100 customers who
• Sample Proportion ( ̂ ) have purchased from Amazon,
an online retail store, on a given
• Sample Standard Deviation ( ) day.

6
 Statistical Inference in Brief

• Statistical Inference is Population Sample

the process of using
data from a sample to Subset

make generalisations
or draw conclusions Inference

about a population. Statistic

Parameter

7
 What is Statistical Inference

• Statistical Inference involves drawing Examples:

conclusions about the properties of a 1. There is a survey about the
population (parameter) based on information federal election. What can we
learn about the outcome of
obtained from a sample (statistic). the real election?
• It distinguishes from descriptive statistics by 2. 100 volunteers’ results from a
clinical trial for a new vaccine
its focus on using the sample information to are observed. What is the
attempt to learn something about the vaccine’s effectiveness?
underlying population.
8
 Two Important Aspects

• How do we select a Population Sample

sample to represent
the population? Subset

• How can we infer

Inference
about the population
Statistic
parameters based on Parameter
the sample statistics?

9
 Estimator

• A sample is a subset of data drawn from a

Estimator
population. Characteristics of a sample are
An estimator is a statistic
called Statistics.
calculated from a random
E.g.: sample that is used to
• The sample mean, is an estimator of , the population estimate some
mean. characteristic, or

• The sample proportion, ( ̂ ) is an estimator of , the

parameter, of the
population.
population proportion.
• The sample standard deviation, (s) is an estimator of ,
the population standard deviation. 10
 Estimate

• Suppose we want to estimate the average MST

mark of students in QM1 in 2026 S1. Estimate
• Sample mean is the estimator of , the Estimate refers to the
numerical value
population mean.
calculated from the data
• The calculated value from a random sample of using an estimator
20 students (e.g. . ) is the estimate.

11
Sampling and Sampling Methods
 Why a Sample?

• Sampling (i.e. selecting a subset of a

• Cost: It is less expensive and less
whole population) is preferred for many
time-consuming to sample 1,000
reasons. television viewers than 20 million TV
viewers in Australia.
• Gathering data on every single element in
a large population is typically not feasible • Practicality: performing a crash test
on every automobile produced is
due to practicality, and constraints on impractical or impossible.
time, money, and other resources.

13
 Selecting a representative Sample

Characteristics of a good sample:

• Valid inference depends on selecting a
Representative: accurately reflects the
representative sample. characteristics of the population

• The way the sample is selected impacts Random: every member of the
population has an equal chance of being
how representative it is.
selected to minimise selection bias.

Independent: Observations are

independently selected.

Adequate size: large enough to provide

reliable estimates.

14
 Sampling Methods

A sampling method is a procedure or method used to select a subset of

individuals, items, or observations from a population to represent that
population in a study.
The most commonly used sampling methods are:

• Simple random sampling

Key Takeaway:
• Stratified random sampling
• In QMB, we only use samples
• Cluster sampling that are collected using the
simple random sampling
• Multistage Sampling method.
15
 Simple Random Sample

• Random sampling is used to select a

representative sample. Simple Random Sample.
A sampling scheme in which
• A random sample selects observations randomly • Random: ensures that every
and independently, to avoid systematic bias member of the population
has an equal chance of being
towards containing too many elements of selected.
certain types. • Independent: Each individual
is selected independently of
• Statistics from such samples allow valid every other individual.
inference about the population.

16
 Simple Random Sample?

• A Poll was conducted to predict the winners of

the next Australian election. All customers at Simple Random Sample.
A sampling scheme in which
this car dealership are surveyed for their voting • Random: ensures that every
intentions. member of the population
has an equal chance of being
• ABC Radio National asks people to dial in and included.
• Independence: every
vote for a public policy change on healthcare.
individual is chosen (or not
• This article talks about the failure to correctly chosen) independently of
every other individual
predict the outcome of the federal election in
Australia in 2019.
17
 Should we always focus on sample size?

A well-designed small sample is better than

a large, poorly designed one!
• The representativeness matters more
than size.
• Larger samples may help improve
precision, but not bias.
• Poor design leads to misleading results.
• Large samples are not always necessary.
18
Sampling Distributions
 From Samples to Sampling Distribution

• We can take many different random

samples from the same population.
• Each sample gives different values of
a statistic (e.g. mean, variance).
• These values form a distribution.
• We now study the properties of
these sample statistics and what
they mean for statistical inference.
20
 From Samples to Sampling Distribution

• The sampling distribution describes the Sampling Distribution:

distribution of a statistic (e.g. sample The probability distribution
of any statistic calculated
mean, sample proportion) calculated from from a random sample is
all possible random samples of a given size, called a sampling
drawn from the same population. distribution.

21
 Sampling Distribution

• The sampling distribution shows the

possible values a statistic can take and
their probabilities.
• It allows us to quantify the variability
in using a statistic to estimate a
population parameter.

22
 Evaluating Sample Mean
Class test marks
250

Population Mean 15.92

• In 2022, QM1 had 1,368 200 Median 16
students. Their marks for a
mid-semester test were
Number of students
150

recorded.
100
• The (population) mean mark
was 15.92, and the SD was 50

2.63.
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Marks

23
 Evaluating Sample Mean

• Suppose I have a random sample (n=10)

taken from this class and calculated the
sample mean of the test marks, .
• Recall: The sample mean is the estimator,
and 15.9 from my sample is the estimate.

24
 Evaluating Sample Mean
1,368 students:
Mean = 15.92 SD = 2.63

• How could we think about the reliability

of based on 10 as an estimator for
?
• Consider what different values of might
result from taking different samples of …….

size 10.
…….

25
 Evaluating Sample Mean

• Could be always be closer to the population

mean as the one I have here?
• Or can it be far away from the true population
mean?
• Can we express the uncertainty from using
only 10 observations?
• Would increasing produce a better in
some sense?

26
 Sampling Distribution of Sample Mean

• All previous questions can be answered

by investigating the sampling Sampling Distribution of :
distribution of The sampling distribution of is
the probability distribution of
• Recall: is an estimator. all of the possible value of
that could arise from the
• Note: is random (we don’t know the
population.
exact value of until the random
sample is drawn)

27
 Evaluating Sample Mean: Class Exercise

• Do not do this exercise until you are asked

to do it in class.
• Open [Link] available in Canvas.
• Calculate the sample mean of the marks.
• Open the quiz using the link or the QR
code.
[Link]
• Enter the sample mean and the sample SD urses/235754/quizzes/314758
you got.
28
 Evaluating Sample Mean: Class Exercise

• Talk to each other and compare the

sample mean you got with what
they have.
• Are they exactly the same?
• Are they very different?
• Do you now agree that the sample
[Link]
mean is a Random variable? urses/235754/quizzes/314758

29
 Evaluating Sample Mean

• I generated 100,000 samples of size

10 and calculated 100,000 sample
means.
• In the histogram, X axis has the values
of the sample means, Y axis has the
probability for each value range.
• What can you tell from this histogram?

30
 Evaluating Sample Mean

• How do we compare the two distributions

Class test marks

250

200
Number of students

150

100

50

0
1 2 3 4 5 6 7 8 9 10 11
Marks 12 13 14 15 16 17 18 19 20
31
Distribution of marks Distribution of
Sampling distribution of Sample Mean

Here are the statistics I calculated;

what do these reveal?
• Mean ( ) = 15.91841 ( =15.92)

• Sd ( ) = 0.831829 ( = 2.63)

• Proportion of above = 0.4990

• Proportion of within 1 mark of

(| - | 1) = 0.7709
32
 Evaluating Sample Mean

• How if I increased the sample size?

• I generated 100,000 samples of size
20 and calculated 100,000 sample
means.
• Would increasing produce a better
in some sense?

33
Evaluating Sample Mean

34
Sampling distribution of Sample Mean

35
Properties: Sampling Distribution of

1 1 1 1
⋯ Recall the addition rule for mean:
4 4
4 ’s are fixed weights, then
1 1 1
⋯ " # 4 4
Also recall that expected value of a

1 1 1 random variable,
⋯ =

Proof not needed, understanding is important

36
 Mean of the Sampling Distribution

• Read it as: the expected value of the sample

mean is equal to the population mean, or the
mean of the sample mean is the population Sampling Distribution of :
mean.
• Interpretation: the sample mean (as a random
variable) is centered around the true value of
the population mean (very powerful thing).

37
 Unbiased Estimator

• The expected value of the sample mean is

the population mean:
• Therefore is an unbiased estimator of . An estimator is unbiased if
the expected value of that
• An unbiased estimator is one which is not estimator is equal to the
tending to be too large or too small. It is, on population parameter.
average, just right!
• Important: unbiasedness does NOT mean
only " # .
38
 Beyond Unbiasedness: Variance of

• Unbiasedness sounds good, but it is not good

enough. Variance? So far, we know that
• we calculate is a
• We can also evaluate its dispersion from the
random variable.
mean. • is unbiased: on
• If the variance is large, a single is more likely average it is equal to
to be far away from the true mean, . the population
parameter, .
• If the variance is small, a single is more likely
to be close to the true mean,
39
 Evaluating Sample Mean
As we have seen here:

Sd ( ) = 0.8318 Sd ( ) = 0.5854

40
 Variance of the Sample Mean of the
Sampling Distribution
, ,…, are independent with var Recall the addition rule for
variance:
1 1 1 1 4 4
9:; <:; 9:; ⋯
4 ’s are fixed weights, then
var 4 4 7
1 1 1
<:; <:; … <:;" # 4 var" # 24 4 cov" , #
4 var" #
1 1 1
⋯ Each pair =>< , ? 0
because of independence

41
Proof not needed, understanding the concept is important
 Standard Deviation of the Sample
Mean of the Sampling Distribution

9:;

BC B
@A
We call this the Standard Error (s.e.)
or
of the sample mean

42
 Variance of Sample Mean

BC
• 9:; ̅ • Larger samples reduce the
variance; they give a better

 Larger ⇒ smaller variance

chance for sample mean to be
close to the true mean.
 is more precise

⇒ larger variance
• Populations with larger
 Larger variances result in less precise
 is less precise sample means.

43
 In Summary

For a simple random sample , ,…, ,

and <:; , • The sample size does not

affect unbiasedness.
Then, the sample mean , • However, larger samples
reduce the variance.
• (unbiased) • Larger samples give a better
BC
• 9:; ̅
chance for the sample mean
to be close to the true
 Larger ⇒ is more precise mean.
 Larger ⇒ is less precise
44
 Sampling distribution of Sample Mean

• The Histogram with a larger

sample size is narrower.
• It suggests lower variability and
higher precision in estimating
the corresponding population
parameter

45
 Consistency

• As the sample size increases, the sampling

distribution of is more concentrated
around the true mean.
is said to be consistent
• Larger samples give more precise sample because:
means (they vary less from sample to • It is unbiased
• Its variance converges
sample).
towards zero as grows
• Increasing the sample size reduces the
spread of sample means, making them more
tightly clustered around the true mean.
46
 Consistency

• As the sample size goes to infinity,

the sampling distribution of
becomes a spike at the true
population mean.
• This means that all possible
outcomes for will be arbitrarily
close to the true population mean!

47
 In Summary

For a simple random sample , ,…, ,

and <:; ,

Then, the sample mean is H> IJIKL K,

• (unbiased)

BC
• 9:; ̅
 Its variance converges towards zero as grows
48
Central Limit Theorem
 Central Limit Theorem

If , ,…, are are independently taken from a population with

and var

We have derived that

B
and var and, sd

Define Z-score

N mean" # Nμ
M
sd" # σ/√n
50
 Central Limit Theorem

Nμ
M
Central Limit Theorem
σ/√n As increases, the
distribution of M converges
The distribution of M becomes
towards N(0, 1) regardless
approximately standard normal for of the distribution of .
large samples, regardless of the
distribution of .

51
 Central Limit Theorem

Our aim is to use to draw inferences

about the population parameter .

Nμ
M
σ/√n

This provides good foundation for

inference, except is unknown.

52
 Central Limit Theorem

The CLT still works if we replace σ by s, then,

Nμ
When σ is estimated by s, a better
~K T
approximation may be provided
s/√n by the t- distribution instead of
standard normal distribution.

53
 Central Limit Theorem

• If X is normally distributed then,

~K
UTV
W/√X T (exactly t-distributed).

• If X is not normally distributed then,

UTV
~K
a
W/√X T (approximately t-distributed).

54
 Recall:

• The t distribution is very similar to the

normal distribution.
• Shape depends on the degrees of freedom, ν
(reads nu), which is n − 1.
• The t distribution has a (slightly) larger
variance than the standard normal
distribution.
• As n increases, the t and normal distributions
become more and more indistinguishable. 55
Evaluating Sample Proportion
 Evaluating Sample Proportion

• For categorical data, we focus on the Example: Who will run the next election?
proportion of a particular category. • Consider the population of all
Australian voters.
• Select a voter at random and define the
• The population proportion is the random variable :

parameter, and the sample proportion Y is

• 1, if the voter prefers the Labour
party,
the estimator. • 0 otherwise prefers the Liberal
party

1, voter prefers Labour

Z
0, voter prefers Liberal
57
 Evaluating Sample Proportion

1, voter prefers Labour party

Z
0, voter prefers Liberal party

Let’s denote the probability that the

voter prefers Labour, then:

1 And 0 1N

If ^ 0.5 ⟹ Labour wins,

If 0.5 ⟹ Liberal wins
58
 Evaluating Sample Proportion

However, is unobserved.
The sample proportion is the
total number of successes in the
If we have a sample of 900 voters, can we
sample.
say something about the real value of ? U
̂= ,
• a is the number of successes in
We can calculate a sample proportion,
Y, from this sample.
the sample- i.e the number of (a
1)s
• is the sample size.
This gives us an estimate of the population
proportion,
59
 Sampling Distribution of Sample Proportion
Registered Australian Voters

• Imagine taking many random samples

from the population and calculating the
sample proportion ̂ for each sample
(different samples of 900 voters).
• The values of ̂ will vary from sample to …….
sample.
• ̂ is random variable and has a sampling ̂ ̂ ̂ ……. ̂
distribution (similar to ),
60
 Sampling Distribution of Sample Proportion

Sampling Distribution of b :
• Using the properties of expected
value and variance (as we did for ̂ cb

the sample mean), we can "1 N #

determine the mean and variance 9:; ̂ cb

of ̂ .
"1 N #
@A ̂ cb

61
 Sampling Distribution of Sample Proportion

• For large sample sizes, the sampling Sampling Distribution of ̂ :

distribution of ̂ is approximately If sample size is sufficiently large such

normal. that d 5 and "1 N # d 5, then,

• This is a result of the Central Limit The sample proportion ( ̂ # is

approximately normally distributed
Theorem. with mean and variance "1 N #/ .

̂ ~e" , "1 N #/ )

62
 Central Limit Theorem

Even for a variable with 0/1 values, the CLT gives

Sampling Distribution of b :
̂ N mean" ̂ # ̂N
M ̂
sd" ̂ # "1 N #/
cb

9:; ̂ cb "1 N #/

@A ̂ cb
"1 N #/

63
 Sampling Distribution of Sample Proportion

Properties of sampling
• As the sample size increases, the g
distribution of f
" ̂ )=
distribution becomes more
And,
@A" ̂ ) "1 N #/ .
concentrated around the population
proportion, reducing variability
(standard error). For sufficiently large samples,
b ~e" , "1 N #/ )

64
 Properties of Sampling Distribution of b

• We can use ̂ to estimate the population

proportion. Properties of sampling
g
distribution of f
• The sample proportion is an unbiased estimator of " ̂ )=
the population proportion ( ). And,
@A" ̂ ) "1 N #/ .
• The sample proportion is also a consistent
estimator of the population proportion. For sufficiently large samples,
b ~e" , "1 N #/ )
• This means that as the sample size increases, the
sample proportion becomes increasingly close to
the population proportion.
65
 Here is the end of Week 6

Here is what we covered!

Key Words:
• Parameter, Statistic and Estimator.

• Simple Random Sampling and Sampling

distribution.

• Unbiasedness, Consistency and Precision.

• Central limit theorem.

66