HE2020 Survey Methods and Sampling Techniques

(An Introduction)

(A) Why Sample?

(B) Basic Sampling Design Process
(C) Non-probability Sampling
(D) Probability Sampling

(A) Why sample? Why not census?

Time Constraint
(i)

Unemployment rate
_________________________

(ii)

Inflation rate
_________________________

(iii)

Presidential election poll

Hence, we can present the results in ______________.

Example: Taipei mayoral election

Survey outcome: On 11 Nov 2014, Dr. Kos political campaign staff expected that Dr. Ko would get 841,313 votes
based on the voting rate 70% in the coming election.
Fact: On 29 Nov 2014, Dr. Ko won the election and had secured 853,953 votes (voting rate: 70.46%).

Example: Presidential election

Survey outcome: The poll cited by Croatia's state TV said Kolinda won 51.2 per cent of the vote Sunday, while
President Ivo had 48.8 per cent. Pollster Ipsos Puls, whose surveys have been accurate in the past, said the error
margin was about 1 percentage point.
Fact: In Jan 2015, opposition challenger Kolinda has become the first female president of Croatia, winning by the
narrowest of margins. She won with _____ of the vote.
Newspaper examples: (Bloomberg, 1 Jan 2015)
Cleaners, laborers and production and transport operators accounted for the highest numbers of older workers,
according to a survey last year by Singapore-based DBS Bank Ltd. A majority of the elderly who were employed
drew gross monthly incomes of less than S$1,500, it showed.
Only ________ Singapore investors is confident their pension accounts will meet their retirement needs, with 47
percent indicating the savings will be insufficient, according to a survey released in August 2014 by Toronto-based
insurer Manulife Financial Corp.

Financial Constraint
(i)

Product satisfaction ( R&D, Change marketing strategy)

(ii)

Small companies/ Individual investigator

Hence, the cost is much lower.

Example: Quality Assurance testing

(a)

_____________

(b)

_____________

(c)

______________

Access Problem
(i)

Average salary of NTU graduates

(ii)

Average taxi expense per month

(iii)

# of infected people of HIV/ AIDS for every 10,000 persons

Hence, it is more realistic to do a sampling survey instead of census.

Less well known, estimates based on sample surveys are often ___________ than those
based on a census because investigator can be more careful when collecting data. A
complete census often requires a large administrative organization and involves many
persons in the data collection. With the administrative complexity and the pressure to
produce timely estimates, many types of errors can be easily injected into census.
[Sharon, 1999].

When should you sample the entire population?

When the population ___________.
When we do not have constraints. We have _____________.
When we have a __________________.

But the point is that the estimated population value based on a sampling data is not
going to be same as true population value.
The different is known as the bias.

Roughly speaking,
true value = ____________________________________________________

Sampling error is the error in _________. It results from taking one sample instead of
measuring every unit in the population.
o

Ensure the sample adequately represents the entire population; increase the
sample size

Proper and unbiased sampling procedure

There is only one way to eliminate sampling error.

Non-sampling error can be further classified into two types: (a) selection bias, and (b)
measurement error (inaccuracy).

(a) Sample selection bias is due to a biased sampled group.

o

Use a correct sampling method

Correct the /change the weight for different groups

Note that large unrepresentative samples can perform ________ small unrepresentative
samples.
7

(b) Measurement error happens if we dont have good quality data/surveys/questions.

Practical problems!
People _________________ or do not understand the questions.
People forget.
People give different answers to different interviewers. Or people want to ______
the interviewers or say what they think an interviewer wants to hear.
Not clear questions. For example, Do you own a car?

Sampling is not mere substitution of a partial coverage for a total coverage.

Sampling is the science and art of controlling and measuring the reliability of useful
statistical information through the Theory of Probability. [Deming, 1950].
We want to have a good estimate of the parameter of interest.
o Unbiased and small variance
o Less sampling and non sampling errors
o Cost effectiveness

(B)The Sampling Design Process

A.

Define the Target Population based on the Survey Projects

o Target Population: The complete collection of observations we want to study.
But, investigators always draw a sample from a population ______than the target
population.
For example:
(i) Not all potential voters have landed phone line.
(ii) Using phone interviews to have the presidential election survey
(iii) Some may refuse to do the survey.

B.

Determine the Sampling Frame

o Sample Frame: It is a list or a set of directions identifying all the sample units in the
population.
For example: _________________________

Example: How many percentages of the households buy groceries from Sheng Siong last month?

Example: Amazon books (www.amazon.com) summarizes reader reviews of the books it sells.
Persons who want to review a book can submit a review online; Amazon then reports the average
rating from all reader reviews on its website. Describe the target population, sampling frame,
sampling unit

10

C.

Select a Sampling Technique

o Probability or non probability sampling?

D. Determine the Sample Size

o ________

E.

Execute the Sampling Process

o Make questions
o Data collection/Transfer/ Data entry
o Analysis programming/ Statistical analysis
o Report

11

(C) Non-Probability Sampling

The non probability sampling does not involve _________. Therefore, we cannot
infer/extend the results/estimates obtained from the survey to the general population.
This type of sampling does not use random selection methods. Often, interviewers select
the representatives (?) from a convenient way.

Convenience Sampling / Opportunity Sampling / Accidental Sampling

o _________________

Quota Sampling
o The survey is continued until a ________________ of the people, households, firms,
etc.

Snowball Sampling
o The interviewee _______ the next interviewee for investigators.

12

(D) Probability Sampling

Simple _______ Sampling (SRS)
o SRS is the simplest form of probability sample. It is a method of selecting n units out of
N such that every one of the NCn distinct samples has an equal chance of being drawn.

_________ Sampling
o It is a method to select a sample of n units out of N. It takes the first unit at random from
the first k units, and every
unit thereafter.
o Therefore, the selection of first unit determines the whole sample.

________Random Sampling
o Divide the population of N units into L subpopulations (known as strata) of each
consisting of , ,..,
units respectively. The L subpopulations are nonoverlapping, and together they comprise the whole population, so that:
+ +..+
.
o If a SRS is implemented on each subpopulation (stratum), then the whole procedure is
described as Stratified Random Sampling.
o However, we can use different sampling techniques for different subpopulations (strata).

13

 _______ Sampling (One Stage Sampling, Two Stage Sampling, Multistage Sampling)
o Divide the population into L groups (clusters). Instead of drawing individual units from the
whole population, we draw 1 out of L clusters, and collect all the information from the
selected cluster.

Other probability sampling techniques

Panel sampling

14