MKTG3010 Marketing Research

SAMPLING
Last Week
• Measurement and Scaling
• Questionnaire and Form Design
Scale Characteristics
• Description: unique labels or descriptors used to designate each value of
the scale
• E.g., 1. Female, 2. Male; 1=strongly disagree, 2=disagree, 3=neither agree
nor disagree, 4=agree, and 5=strongly agree
• Order: the relative sizes or positions of the descriptors
• “greater than,” “less than,” and “equal to”
• Distance: it means absolute differences b/t the scale descriptors are
know and expressed in units
• E.g., # of persons living in your household ______
• Origin: it means the scale has a unique or fixed beginning or true zero
point
• E.g., What is the annual income of your household?
Primary Scales of Measurement
Scaling Techniques
Key Issues in Questionnaire Design
• For each question, think about:
• Is the question necessary? How will I use the data?
• Are several questions needed instead of one?
• Is the respondent able to answer the question?
• Is the respondent informed (filter questions, “don’t know” option)
• Can the respondent remember
• Can the respondent articulate
• Is the respondent willing to answer the question?
• Minimize effort, manipulate the context when necessary, legitimize the
purpose, what for sensitive topics
• Ultimate goal: tap respondents’ true responses w/o influencing them
The Do’s and Do Not’s of Question Wording
1. The question should be focused on a single issue
2. The question should be brief and easy to understand
3. The question should be interpreted the same way by all respondents,
avoid ambiguous wording
4. Avoid leading/biasing questions
5. Avoid implicit alternatives
6. Avoid implicit assumptions
7. Avoid generalizations and estimates
8. Have both positive and negative statements
9. Be alert to sensitive questions
Next Week
• Submit your questionnaire
• Submit Assignment 1
Research Design: Components
• Define the information needed
• Design the exploratory, descriptive, and/or causal phases of the
research
• Specify the measurement and scaling procedures
• Construct and pretest a questionnaire (interviewing form) or an
appropriate form for data collection
• Specify the sampling process and sample size
• Develop a plan of data analysis
Why sampling is important?
Consider 2016 US Election Polling:
• New York Times: Hillary Clinton has an 85% chance to win.
• FiveThirtyEight: Hillary Clinton has a 71% chance to win.
• Princeton Election Consortium: Hillary Clinton has a 98-99%
chance to win.
Sampling
• Population: who you want to learn about, e.g. HK commuters
• Doesn’t have to be people (e.g. could be stores)
• Sample plan: how each sample element, or unit, is to be drawn
from the total population
• Goal: representativeness!
• In what way has technology made it easier or harder to draw a
representative sample?
• Sample size: how many population units to include in the sample
• Goal: accuracy!
Census versus Sample

• Census: A complete enumeration

of a population or study subjects.
• HK census: every ten years
• All households

• Sample: A subgroup of the

population selected for
participation in the study.
When is sample appropriate?
• Population size is large
• Both cost and time associated with obtaining information from
the population is high
• Quick decision is needed
• To increase response quality since more time can be spent on
each interview
• Population being dealt with is homogeneous
• If census is impossible
Potential Sources of Error
Errors Associated with Sampling
Sample Frame Error

• The sample frame shape does not take in all of

the population shape.
• The sample frame takes in some area that is
outside the population’s boundary.
• Use telephone users as sample frame of election
voters:
• Not all voters users telephone
• Some telephone users are not voters (e.g.,
foreigners)
Example: Landline vs. Mobile Phone Sample Frame
• [Link]
Random Sample Error
• Sampling Method
• Sampling Size
• We will cover more about it later
Non-response error

those who responded

Response rate =
those contacted
Sample Design Process
Define the Target Population

Determine the Sampling Frame

Select Sampling Technique(s)

Determine the Sample Size

Execute the Sampling Process

1. Define the Target Population
• The target population is the collection of elements or objects that
possess the information sought by the researcher and about
which inferences are to be made.
• Who do you want to learn about?
• The target population should be defined in terms of
• Elements—object about which or from which the info is desired (e.g.,
respondent)
• Sampling units—element, or entity contains the element (e.g.,
household)
• Extent—geographical boundaries
• Time—the time period under consideration
Examples

• Election polling
• A pharmaceutical firm offering a new drug
• CUHK Bookstore designing an advertising campaign

• HK MTR designing its new station

2. Determine the Sampling Frame
• A sampling frame is a representation of the elements of the target
population. It consists of a list or set of directions for identifying
the target population.
• How can you get access to them?
• e.g.
• A telephone book
• A mailing list purchased from a commercial organization
• A city directory
• Yellow page
Examples

• Election polling
• A pharmaceutical firm offering a new drug

• CUHK Bookstore designing an advertising campaign

• HK MTR designing its new station

3. Select Sampling Technique(s)

Sampling
Techniques

Nonprobability Probability
Sampling Sampling
Techniques Techniques
Sampling Techniques
The ultimate goal is to find out … of the population. Having a
representative sample is soooo important!
• Nonprobability sampling: relies on convenience or the personal
judgement of the researcher
• “May” yield good estimates
• Not statistically projectable to the population
• Probability sampling: sampling units are selected by chance
• Each element has a fixed chance of being selected
• Allow inferences or projections about the target population
Nonprobability Sampling Techniques

Nonprobability Sampling Techniques

Convenience Judgmental Quota Snowball

Sampling Sampling Sampling Sampling
Convenience and Judgmental Sampling
• Convenience sampling
• The researcher decides the selection of sampling units
• Least expensive and least time-consuming, very convenient
• e.g., “people on the street” interviews

• Judgmental sampling
• The researcher chooses the sampling elements because she or he
believes they represent the population of interest.
• The researcher acts as an “expert”.
Quota Sampling
• Aka two-stage restricted judgmental sampling
• Step 1:
• The population is divided into several categories.
• E.g., male and female, 18-35 years old and above 35, White and
Hispanic,…
• Assigning quotas to each category.
• The quota can be based on some findings (e.g., census).
• E.g., 50 males and 50 females, 80 White and 20 Black, …
• Step 2: Select elements in categories based on convenience or
judgment until quota is reached.
Quota Sampling
Snowball Sampling
1. An initial group of respondents is selected, usually at random.

2. These respondents are asked to identify others who belong to

the target population of interest.

3. Continue sampling based on the step 2 results.

Exercise: Which sampling technique is used?
Exercise: Which sampling technique is used?
Exercise: Which sampling technique is used?

Snowball Sampling!
Exercise: Which sampling technique is used?
Exercise: Which sampling technique is used?
Probability Sampling Techniques

Probability Sampling Techniques

Simple Random Systematic Stratified Cluster

Sampling Sampling Sampling Sampling
Simple Random Sampling
• Each element of the population is sampled with equal probability.
• RDD: Random Digit Dialing
Is Random Sampling “Random Enough?”
• Can get outcomes that don’t “seem” random

•
Or

• You may select a row who are similar and not representative
Systematic Sampling
• The sample is chosen by selecting a random starting point and
then picking every 𝑖th element in succession from the sampling
frame.
• E.g., in a telephone interview, select all telephone numbers
ending with the same four digits 0189.
Stratified Sampling
• A two-step process
• The population is divided into subgroups called strata.
• e.g., gender, age
• Elements are then randomly selected from each stratum.
Cluster Sampling
1. Divided the target population into
mutually exclusive and collectively
exhaustive subpopulations, or
clusters.
• E.g., states, postcode, colleges
2. A random sample of clusters is
selected based on a probability
sampling technique, such as SRS.
3. All or a random sample of elements
are selected from the chosen
clusters
Stratified vs. Cluster Sampling

Stratified Sampling Cluster Sampling

•Homogeneity within group •Homogeneity between groups
– Ex.: all males are same; all – Ex.: Boston inner city
females are same. neighborhoods are the same.
•Heterogeneity between groups •Heterogeneity within group
– Ex: males and females are – Ex: residents within a
different. neighborhood differ.
•All groups are included •Random selection of groups
– Ex.: include males and – Ex: randomly select
females in sample. neighborhood and go door-
to-door
Stratified vs. Cluster Sampling
• Example: Sample HK college students

• Stratified Sampling:
• Divide population into freshman, sophomore, junior and senior
• Randomly sample 1000 students from each group.

• Cluster Sampling:
• Divide population into CUHK, HKU, HKUST, Poly U, etc.
• Randomly sample 2000 students from CUHK and 2000 from HKU.
Exercise: Which sampling technique is used?
Exercise: Which sampling technique is used?
Exercise: Which sampling technique is used?
Exercise: Which sampling technique is used?
Exercise: Which sampling technique is used?

Stratified Sampling!
More Descriptions
• [Link]
Nonprobability vs. Probability Sampling
Errors Associated with Sampling
Non-Response Problem
• Respondents may:
1. Refuse to respond
2. Lack the ability to respond
3. Be inaccessible
• Sample size has to be large enough to allow for non response
• Those who respond may differ from non respondents in a
meaningful way, creating biases
• Seriousness of non-response bias depends on extent of non
response
Solutions to Non-response problem
• Improve research design to reduce the number of non-responses

• Repeat the contact one or more times (call back) to try to reduce
non-responses

• Attempt to estimate the non-response bias

Sample Design Process
Define the Target Population

Determine the Sampling Frame

Select Sampling Technique(s)

Determine the Sample Size

Execute the Sampling Process

4. Determine the Sample Size
• Three Axioms
• The only perfectly accurate sample is a census.
• A random sample will always have some inaccuracy, which is referred
to as “sample error.”
• The larger a random sample is, the more accurate it is, meaning the
less sample error it has.

• Basis of statistics: Law of Large Number

Example
• We want to know the mean income of all HK households.
• ~ 2.5M households and a census is too expensive.
• We randomly sampled 𝑛 = 1,000 households, and their incomes
are 𝑋1 , 𝑋2 , … , 𝑋𝑛 .
𝑋 +⋯+𝑋𝑛
• The mean income of the sampled households is 𝑋ത = 1
𝑛
ത
• But what is the mean income of all HK households? It’s not 𝑋!
• “The mean of the sample is NOT equal to the mean of the whole
population!”
Our Task
• We want to know the mean of the population.

• However, we can only directly measure the mean of the sample.

• We want to infer the mean of the population from the mean of

the sample.
Statistics Review
• Mean

• Standard deviation

• Standard error of the mean 𝑋ത

Statistics Review
• Confidence Interval for the mean of the population

[𝑋ത − 𝑧𝜎𝑥ҧ , 𝑋ത + 𝑧𝜎𝑥ҧ ]

• The mean of the population is around the mean of the sample.

• z value: A number associated with confidence level (CL)
• For confidence level 95%, the z value is 1.96
Interpretation
• We sampled 1,000 HK households
• The mean income of sampled households is 𝑋ത = 30,000.
• The standard error is 𝜎𝑥ҧ = 1, 000.
• The z value for 95% confidence level is 1.96.
• 𝑋ത − 𝑧𝜎𝑥ҧ = 28,040 and 𝑋ത + 𝑧𝜎𝑥ҧ = 31,960
• How to interpret this?
• With probability 95%, the mean HK household income is between
28, 040 and 31, 960
• [28040, 31960] is called confidence interval here.
Precision
• With probability 95%, the mean HK household income is between
30,000 – 1,960 = 28,040 and 30,000 + 1,960 = 31,960

• D = 𝑧𝜎𝑥ҧ = 31960 – 30000 = 30000 – 28040 = 1960 is called the

precision.
• Confidence interval is also written as [𝑋ത − 𝐷, 𝑋ത + 𝐷].
• We want D as small as possible!
Our Problem
• We already answered the question “how to interpret the result?”

Sampling N people Result of some precision

• Now the new question is, “how many individuals to sample to

get the desired result”?
1. What is the desired result?
2. How many individuals to sample to get the it?
How many individuals to sample?
1. Specify the level of precision.
2. Specify the level of confidence.
3. Determine the z value associated with the confidence level.
4. Determine the standard deviation of the population.
5. Determine the sample size using the formula for the standard
error of the mean.
Example
• Purpose: We want to know the average height of HK males.

• How many HK males to sample?

How many individuals to sample?
1. Specify the level of precision.
• Recall that precision is D, e.g., D = 2cm.

2. Specify the level of confidence.

• Confidence level is CL, e.g., CL = 95%.

Then, the sample results should be something like “The mean height
of Hong Kong males is [M – 2cm, M + 2cm] with confidence level 95%.”
Here M is something we obtain from the data.
How many individuals to sample?
1. Specify the level of precision, e.g., D = 2cm.
2. Specify the level of confidence, e.g., CL = 95%.
3. Determine the z value associated with the confidence level.
• When CL = 95%, the associated z value is 1.96.
How many individuals to sample?
1. Specify the level of precision, e.g., D = 2 cm.
2. Specify the level of confidence, e.g., CL = 95%.
3. Determine the z value associated with the confidence level, e.g.,
z=1.96.
4. Determine the standard deviation of the population.
• There is NO standard way doing this: we do not have data yet.
• Can be estimated from pilot-study data, based on the researcher’s
judgment, inferred from the previous survey data.
• E.g., we asked a medical examiner who suggested that the standard
deviation of HK males is around 𝜎 = 10 cm.
How many individuals to sample?
1. Specify the level of precision, e.g., D = 2.
2. Specify the level of confidence, e.g., CL = 95%.
3. Determine the z value associated with the confidence level, e.g.,
z = 1.96.
4. Determine the standard deviation of the population, e.g., 𝜎 2 =
100.
5. Determine the sample size using the formula for the standard
error. 𝜎2 𝑧 2 102 × 1.962
𝑛= 2 𝑛= 2 = 96
𝐷 2
We need to sample 96 individuals!
 MKTG3010 Marketing Research

R
Data Entry
Some Data Checks
Read and Write CSVs
Load Data
• Use “/” in filepath
table()
• table() for categorical variable

• The counts can be converted to proportions with [Link]()

Two-way Tables

• Tables index [row, column] like most things in R

Descriptive Functions
Percentile (Quantile) Function
Summary of DataFrame
Summary of DataFrame Elements
Exercises
1. Load “[Link]”
2. Compute a frequency table for the country of origin (country) for
site visits. After the United States, which country had the most
visitors?
3. Compute a two-way frequency table for the intent to purchase
(intentWasPlanningToBuy), broken out by user profile (profile)
4. What are the proportions of parents who intended to purchase?
The proportions of teachers who did?
Next Week
• Data Preparation
• Conjoint Analysis
• Data Visualization