Statistics 1
Chapter 10:
Sampling Design
Statistics for Business and Economics
Section 6.1 and Chapter 17
Dr Joan Gan
Dr Raymond Teo
Sampling vs Census
Census
Obtaining information from everyone in the population
Sampling
Obtaining information from a smaller group and making
inference on a larger population
Sampling Design 2
1
Statistics 1
Why may a sample be preferred to a census?
1. Conducting a study on a sample is cheaper.
2. Information can be collected faster with a sample. In
a census, the collection may take so long that by the
time results are obtained, it may not be of use to the
decision maker.
3. Results from sampling can be used to make inference
about the population.
4. Sometimes, census may introduce non-sampling errors.
Sampling Design 3
Types of sample
Sampling Design
4
2
Statistics 1
Types of Sampling Methods
Random Sampling Non–Random Sampling
Methods Methods
(The probability of each (The probability of each
unit being chosen is unit being chosen is
known) unknown)
1. Simple random 1. Convenience sampling
sampling 2. Judgment sampling
2. Stratified sampling 3. Quota sampling
3. Cluster sampling
Sampling Design 5
Types of Sampling Methods
Random Sampling Non - Random Sampling
Methods Methods
This approach requires No sampling frame is
the use of a sampling required here.
frame
A sampling frame is a list of all members of a population
Sampling Design 6
3
Statistics 1
Random sampling techniques have the
following advantages
• Avoids systematic bias
• Allows assessment of size of sampling error
(covered in earlier chapters)
Sampling Design 7
Probability sampling
methods
Sampling Design 8
4
Statistics 1
Simple Random Sampling
Each individual is chosen entirely by chance and
each member of the population has an equal
chance of being included in the sample.
• This method is acceptable when the population is
homogeneous.
• Simplicity of the method is a plus point
Sampling Design 9
Random Number Table
56976 10799 52098 04184 54967 72938
83125 85077 60490 44369 66130 72936
55503 21383 02464 26141 68779 66388
47019 06683 33203 29603 54553 25971
84828 61152 79526 29554 84580 37859
08021 31331 79227 05748 51276 57143
36458 28285 30424 98420 72925 40729
05752 96065 36847 87729 81679 59126
26768 02513 58454 56958 20575 76746
Sampling Design 10
5
Statistics 1
Stratified Random Sampling
There may be factors that divide the population into
sub-populations. This has to be accounted for when
we select a sample from the population in order that
it is representative of the population. This is achieved
through stratified sampling.
Stratified sampling increases the precision of an
estimate and in so doing, reduces the standard error.
Sampling Design 11
Situations where stratified sampling is useful
Stratified is particularly useful when the population is
heterogeneous, or dissimilar.
It may also be useful when estimates of the population
may be needed for each strata.
E.g. when you need to know behaviour difference
between part-time and full-time students at SIM.
It is also possible that the sampling frame (lists) are
separate when obtained. E.g. SIM may have a
separate list of local students and International students.
Sampling Design 12
6
Statistics 1
Details on technique
Population divided into strata
Each strata to be homogeneous in terms of some
characteristic
Within each stratum, a sample of elements are
selected
Aggregate all sample elements to form stratified
random sample
Proportionality to population to be considered
Sampling Design 13
Systematic random Sampling
a 1-in-x systematic random sample is obtained by
randomly selecting one of the first x units in the
sampling frame and then selecting every
subsequent x-th unit.
• Very easy to implement but needs to be careful of
periodic property in the data to be collected
Sampling Design 14
7
Statistics 1
Cluster Sampling
Cluster sampling is a sampling technique where the
entire population is divided into groups, or clusters,
and a random sample of these clusters are selected.
In a ONE-STAGE CLUSTER SAMPLE, all observations
in the selected clusters are included in the sample.
Sampling Design 15
Cluster Sampling - Multi-stage design
Multistage design in cluster sampling combines
various sampling e.g. a cluster sampling following by
a subsample taken.
Sampling Design 16
8
Statistics 1
Situations where cluster sampling is useful
1. It is the preferred method when a random sample
would produce a list of subjects so widely scattered
that surveying them would prove to be far too
expensive, for example, people who live in different
parts of China. Compared to simple random
sampling, it is practical and more economical.
It is not a problem when mail surveys or telephone surveys are
used. It is a major problem when personal interview is chosen
as the contact method over a large geographical region. (You
would not use a cluster sampling method in Singapore.)
The cost would make a simple random sampling method
impossible.
Sampling Design 17
2. When sampling frame of whole population is
not immediately available.
3. You will be able to complete a larger number
of interviews.
4. You will use the method if cluster information is
important to you. e.g. a district in China.
Sampling Design 18
9
Statistics 1
Weaknesses
There is a trade-off in accuracy as the clusters may
not be representative of the population e.g. China.
In other words, the variance of the sample estimate
is greater.
Sampling Design 19
Multistage Sampling
refers to the case when sample selection occurs at
two or more successive stages
e.g. You could also use a stratified sampling of clusters to
improve representativeness
Sampling Design 20
10
Statistics 1
Non-Probability sampling
methods
Sampling Design 21
Convenience Sampling
Individuals are selected from the members of the
population which happen to be in the vicinity.
e.g. those people in a shopping mall at the time of the
survey, say a school day.
Sample is unlikely to be representative of the population
Sampling Design 22
11
Statistics 1
Judgement Sampling
Uses expert opinion to judge which individuals to
choose.
e.g. more emphasis given to the representation of the
sample through judgement.
Sampling Design 23
Quota Sampling
In quota sampling, the selection of the sample is
made by the interviewer, who has been given quotas
to fill from specified sub-groups of the population.
Sampling Design 24
12
Statistics 1
Quota Sampling Table (by percentage)
Quota Sampling Table (by number)
Sampling Design 25
Weaknesses of quota sampling
However, interviewer selection bias can result.
e.g. when they avoid respondents who looks
threatening.
As quota sampling is a non-probability sampling
method, sampling error cannot be estimated.
Therefore, we cannot apply the basic rules of
inference.
Sampling Design 26
13
Statistics 1
Arguments for Quota Sampling
• Quota sampling is relatively cheaper (about
one-third the price of a random interview).
However it may get costly if the controls are
increased.
• Survey can be done quickly.
• When sampling frame is not available.
Sampling Design 27
Therefore, you will use quota sampling when:
When time is limited.
Some common quotas include age, gender,
socio-economic groupings
When a sampling frame is not available
Quota sampling is a non-probability sampling
method, and therefore, a sampling frame is
not required.
When there is a tight research budget
Sampling Design 28
14
Statistics 1
When accuracy is not important
When you only want to have a ‘feel’ of things
Sampling Design 29
An Area of Caution
Omission of non-respondents
This is an issue when the non-respondents
are not missing at random i.e. it is biased
towards a particular group.
Sampling Design 30
15
Statistics 1
Think:
You have been asked to make a sample survey of each of the
following. Would you use random or quota sampling? Explain.
a) Airline pilots, for their company, about their use of holiday
entitlement in order to bring in a new scheme.
b) Possible tourists, about their holiday destinations and the likely
length of time and money they expect to spend on holiday in
the next year, for a holiday company planning its holiday
schedule and brochure for next year.
Sampling Design 31
Think:
What would you use for a sampling frame if:
a) You wanted to determine the percentage of SIM students who
drive to SIM.
b) You wanted to find out the percentage of TV viewers who
watched the American Idol?
Sampling Design 32
16
Statistics 1
Errors occurring in
surveys
Sampling Design
33
TYPES OF ERRORS
Error Types
• Occurs as a Non
Sampling
result of sample sampling
error
selection, rather error
than performing a
censes (margin of error)
Selection Response
Bias Bias
Sampling Design 34
17
Statistics 1
TYPES OF ERRORS
Error Types
Non
Sampling
sampling
error
error
Wrong actual
• Inadequate sampling measurements due to:
frame • Not understanding
• Sampling frame not strictly Selection Response words in
adhered to Bias Bias questionnaire
• Non-response bias • Interviewer bias
Sampling Design 35
Sampling errors can be quantified in advance.
(Well, by now, you can relate sample size to
confidence levels, can’t you?)
Non-sampling errors are errors that do not
result from the sampling procedure being used.
They are difficult to detect and hence not
possible to quantify.
Sampling Design 36
18
Statistics 1
Pilot and post-enumeration
surveys
Sampling Design
37
Pilot Survey
A pilot survey is a small scale replica of the
actual survey.
Note how the pilot survey addresses some of the errors
we saw previously:
• It helps you determine the errors attached to different
kinds of questions and hence underpin the design chosen.
• It helps you sort out non-sampling errors:
Do people understand the questionnaire?
Are the interviewers working well?
Questionnaire organisation issues
Sampling Design 38
19
Statistics 1
Post Enumeration Surveys
The post-enumeration survey (PES) is a method for
evaluating the results of a census. It is important to
examine the quality and limitations of census data
and to understand the types and extent of
inaccuracies that occur. A sub-sample may be
re-interviewed to make sure they understood the
questions and answered correctly. It is also used
to test the consistency of replies.
Hence, the PES allows census organizations to
uncover deficiencies in the methodology of the
census and make adjustments for future censuses.
Sampling Design 39
Non-response and
response bias
Issue areas
Sampling Design
40
20
Statistics 1
Non-Response Bias
To see if non-response is a problem (i.e. will
bias the results), you might want to compare
between those who responded and those who
did not:
• You can compare their demographic
characteristics or socio-economic
characteristics.
.
Sampling Design 41
In short, non-response would not be a problem
if those who choose not to respond are random.
If they are not random e.g. those who are very
rich etc, you have a problem.
Sampling Design 42
21
Statistics 1
Two kinds of non-responses:
• Item non-responses
When a sampled member fails to
respond to the question in the questionnaire
• Unit non-responses
When no information is collected from a
sample member
Sampling Design 43
Dealing with non-responses
• Interviewer training
• Improving survey procedures
• Doing call backs
• Increasing incentives
Sampling Design 44
22
Statistics 1
Response Bias
Response errors are errors in responses. This
may arise from a misunderstanding of the question
that was asked or wish to deceive.
Some sources of errors:
- Interviewer related
- Respondent related
Sampling Design 45
Further Illustration
Question Misunderstanding
• This is quite an issue when you are doing a study
in different countries. I did one in China
(questionnaire was translated into Chinese). I
intended to find out the desirability of a product
to the Chinese (from the perspective of buying it).
It was misinterpreted as owning it. It turned
out that the more expensive the product, the
more willing they are to pay for it. Ridiculous,
isn’t it?
Sampling Design 46
23
Statistics 1
Further Illustration
Being Untruthful
• How would you answer to these questions?
1) How much alcohol do you consume a week?
2) Do you have extra-marital affair?
3) Have you cheated in a test before?
Some questions may elicit untruthful responses
from respondents. You need to know which
questions these are.
Sampling Design 47
Reducing Response Bias:
• Recruiting better interviewers
• Training and close supervision of interviewers
• Re-interviewing of respondents
• Doing post-enumeration surveys
Sampling Design 48
24
Statistics 1
Contact Methods
Sampling Design
49
Personal Interviews
Advantages
Good when questionnaire is long, when details
need to be probed, when explaining difficult
concepts (High flexibility)
Exhibits can be shown (samples or product taste
tests)
Response rate is higher
Sampling Design 50
25
Statistics 1
Disadvantages
High costs
Longer time
Anonymity
Interviewer cheating
Field control
Sampling Design 51
Telephone Interviews
Advantages Disadvantages
Faster results No exhibits
Cheaper Length of Interview
Wide geographical Not everyone has a
coverage telephone
Access hard-to-reach
people
Privacy
Better control
Sampling Design 52
26
Statistics 1
Postal or Mail Surveys
Advantages Disadvantages
Wide geographical Mailing lists difficult to
coverage obtain
Thoughtful answers Unidentifiable respondents
Sensitive questions Questionnaire exposure
No interviewer bias No interviewer assistance
Cheap Assume literacy
No control problems Low response rates
Sampling Design 53
Asian Women’s Reasons for Wearing Figure-Revealing Clothes
COUNTRY IT MAKES ME IT MAKES MY IT LOOKS I FEEL
LOOK BODY LOOK GOOD ON ME CONFIDENT
FEMININE(%) MORE (%) ABOUT MY
BEAUTIFUL (%) BODY (%)
Beijing 53.7 68.3 48.8 31.7
Seoul 33.3 20.0 46.7 13.3
Toyko 52.2 47.8 17.4 4.3
Hong Kong 43.8 68.8 31.3 12.5
Bangkok 28.6 14.3 50.0 50.0
Jakarta 66.7 33.3 33.3 44.4
Source: Asian Women Now (Wacoal Human Science Research Centre, 1993)
Sampling Design 54
27
Statistics 1
Examiners will look for an intelligent justification of
your decisions.
What would you choose as the contact method
for each of the following?
a) A survey of shopping patterns
You might need to use a face to face convenience
sampling at the shopping centre. As the survey is
not of a very important nature, mail surveys would
elicit a very low response rate. Using a telephone
interview would exclude tourists whom you cannot
get assess to.
Sampling Design 55
b) A survey of businessmen and their attitude to
a new item of office equipment.
A telephone interview is appropriate since all of
them will have a telephone, and the questions
are simple to ask.
Here, booking time for a telephone interview at
work should be much more effective than waiting
for a form to be filled in or even sending
interviewers to disrupt office routine.
Sampling Design 56
28
Statistics 1
c) A survey of teachers about their pay and
conditions
Here, on the spot interviews will not elicit the
level of detail needed. This is because most
people do not remember their exact pay
particularly if you need several years’ information.
We need to give the respondents some time
to think, and if they need to obtain previous
records, they should be given the opportunity to.
Hence, a postal or mail survey should be the
choice of contact method.
Sampling Design 57
29