Principles of Marketing

Report
MKGT 368
Review for Exam 1
Spring 2011
Role of
Marketing Research
in Managerial Decision-Making
Chapter 1
What is Marketing?
• American Marketing Association Definition:
• Marketing is an organizational function and a set of processes for
creating, communicating, and delivering value to customers and for
managing customer relationships in ways that benefit the organization
and its stakeholders.
• In sum, marketing is about…
• meeting needs
• delivering value to all people affected by a transaction
• getting the right product to the right folks at the right time/place for the
right price using an appropriate combination of promotional techniques
(the four Ps)
What is Marketing Research?
• American Marketing Association (p. 4 in your book):
• …the function that links an organization to its market through the
gathering of information. This information allows for the identification
and definition of market-driven opportunities and problems and allows
for the generation, refinement and evaluation of marketing actions. It
allows for the monitoring of marketing performance and improved
understanding of marketing as a business process.
• Malhotra & Peterson (2006, p. 5):
• …the systematic and objective identification, collection, analysis,
dissemination, and use of information that is undertaken to improve
decision making related to identifying and solving problems (also known
as opportunities) in marketing.
• Feinberg et al. (2008, p. 4):
• … the systematic process of using formal research and consistent data
gathering to improve the marketing function within an organization.
This information is used to identify opportunities and problems,
monitor performance, and link marketing inputs with outputs of
interest, such as awareness, satisfaction, sales, share and profitability.
The “Marketing Concept”
• Need for marketing research based on “marketing concept”
• Idea introduced in 1952, GE’s Annual Report:
• The (marketing) concept introduces the marketer at the beginning
rather than at the end of the production cycle and integrates
marketing into each phase of the business. Thus, marketing,
through its studies and research, will establish for the engineer, the
designer, and manufacturer, what the customer wants in a given
product, what price he (or she) is willing to pay, and where and
when it will be wanted. Marketing will have authority in product
planning, production scheduling, and inventory control, as well as
in sales, distribution, and servicing of the product.
• Gave rise to the “Marketing System”
• Conceptual model linking Independent Variables (causes) to
Dependent Variables (outcomes)
• Understanding the link between IVs and DVs (and reducing
uncertainty) is a key function of marketing research 
Marketing System
Independent Variables
Marketing Mix
(controllable)
Pricing
Promotion
Product
Distribution
Situational Factors
(uncontrollable)
Demand
Competition
Legal/political
Economic climate
Technology
Gov regulation
Dependent Variables
Understanding relationship
between IVs and DVs
is a key function of MR
Behavior
Awareness
Knowledge
Liking
Preference
Intent to buy
Purchase
Performance
Measures
Sales
Market share
Profit
ROI
Image
From Feinberg et al. (2008)
The Decision-Making Process
1. Recognize a unique marketing problem or opportunity
2. Clarify the decision (what do we need to know?)
3. Identify alternative courses of action
4. Evaluate the alternatives
5. Select a course of action
6. Implement selected course of action and monitor results
From Feinberg et al. (2008)
Common Questions Addressed
by Marketing Researchers
•
Where are new market opportunities (based on macroenvironmental trends)?
•
How should we segment the market (based on customer characteristics)?
•
How are we doing (compared to the competition)? Are consumers satisfied with our
product or service? If not, what should we improve?
•
How should we position our product (relative to the competition)?
•
How will people respond to a new product concept? Test marketing…
•
If our product is priced at $100, what will be the expected demand?
•
How effective is our advertising? Promotions? Sales force?
•
What’s in store for the future, and how should we adapt?
Marketing Research Process:
Transforming Data into Information
Chapter 2
Overview
• Types of Marketing Research Firms
• When is Marketing Research Needed?
• Decision-Makers vs. Researchers
• Iceberg Principle: Symptoms vs. Underlying Problems
• Steps in Marketing Research
• Elements in a Marketing Research Proposal
• Unethical Activities in Marketing Research
Marketing Research Industry
Research Supplier
Internal
External
Full Service
Syndicated
AC Nielsen
Customized
Synovate
Limited Service
Internet
Greenfield
On-Line
Field
Services
Data Coding
and Entry
Data
Analysis
Field Work
Chicago
Davis
Coding
Group
SDR
Atlanta
Malhotra & Peterson (2006)
Decision
Maker
Exhibit 2.3
When is Marketing Research Needed?
Type of
information
Nature of
decision
Can decision problem be resolved
with subjective information?
NO
Is problem of strategic importance?
YES
NO
Don’t undertake the
Info research process
Bring in
Marketing Researcher
YES
Availability
of data
Is secondary data inadequate for
addressing the problem?
YES
NO
Time
constraints
Is there enough time to collect
data for managerial decision?
YES
Are there enough resources
($, people) to carry out the study?
NO
Resources
required
NO
YES
Cost/Benefit
Ratio
Does value of research
outweigh costs of research?
YES
NO
Do undertake the
Info research process
When NOT to conduct research…
1. Sufficient information for a decision already exists
2. Insufficient time for research – must make an immediate decision
3. Insufficient resources for research
4. When costs of research are greater than its benefits
Components of the Research Proposal
1. Purpose of proposed research plan (problem, objectives)
2. Type of study (e.g., exploratory, causal, primary, secondary etc.)
3. Define target population and sample size
4. Describe sampling technique and actual data collection methods to be used
5. Research instruments to be used
6. Possible managerial benefits
7. Proposed cost of whole project
8. Describe primary researchers and research firm
9. Proposed tables (how data might be presented)
Researchers vs. Decision-Makers
Researchers
Decision-Makers
• Like to explore new questions
• Want info to confirm decision
• Can tolerate long investigations
• Want quick information
• Not concerned about cost
• Less willing to pay for more info
• Enjoy surprises
• Dislike & reject surprises
• Tentative; speak in probabilities
• Decision- and results-oriented
• Interested in past behavior
• Interested in future performance
Iceberg Principle: Symptoms vs. Problems
Four Broad Phases in Information Research
Ten steps 
Ten Steps in Information Research
Unethical Activities…
•
by Client (End User)
• Solicit proposals, but choose none. Use proposals as a guideline for how to
conduct one’s own study.
• Promise a long-term relationship to get a low introductory rate, but then
never follow through with more projects
•
by Researcher
• Unethical pricing: promise low price, then jack it up
• Fail to provide (promised) incentives to research subjects
• Abuse respondents; promise short survey that turns into an hour; pass along
information without permission; collect information without permission
• Selling useless research services
• Interviewers make up data (“curbstoning” or “rocking chair” interviewing)
• Interviewers create “phantom” data (duplicate actual data to boost sample)
• Change or fail to report results in an effort to reach a certain conclusion
•
by Respondent
• Give misleading responses (can include “socially desirable” responding)
Qualitative Methods:
Interviews and Focus Groups
Chapters 6-7
Qualitative vs. Quantitative Methods
• Qualitative
• Used in exploratory designs to gain prelminary insights into
decision problems and opportunities
• Quantitative
• Using formalized standard questions and predetermined
response options (yes, no) in questionnaires or surveys
administered to large numbers of respondents
• Differences Between Qualitative and Quantitative Approaches 
Focus Groups
• Focus Groups
• Formalized small group of people have an interactive,
spontaneous discussion on one topic or concept
• Can…
• help identify root problem underlying symptoms
• help identify questions to ask in a survey
• provide insights into quantitative results
• uncover hidden needs, wants, attitudes, feelings,
perceptions and motives regarding products/services
• lead to new ideas for products/services
• help develop new measures for quantitative survey
• provide insights into how people “experience”
products/services (what they mean to them)
Composing a Focus Group
• Selecting Participants
• Select a good group of participants (relatively homogenous groups make
people feel comfortable, but should have some variability in views)
• Potential group members should have enough knowledge to contribute
• Try to incorporate some randomization in selection (within a target group)
• Size should be between 8-12 people with a moderator
• Use a friendly invite and provide incentives (typically between $75-100)
• Pick a comfortable location
Some Additional Interview Techniques
• Case Study
• Analyze in depth one or more situations similar to the problem
you are trying to solve
• Experience Interviews
• Interview people believed to be knowledgeable about the
problem you are trying to solve
• Protocol Interviews
• Ask people to verbalize the thought processes and activities
they would go through in a given situation (e.g., buying a car)
• Articulative Interviews
• Listening to people in order to identify value conflicts they may
have (e.g., want to buy a nice bike but also be frugal)
Projective Techniques
• Definition
• Techniques that allow a person to “project” their thoughts, feelings
or motives onto others, a situation, or an object
• Types
• Word association: When I say “GO” you say_____
• Sentence completion: Students at WSU are _____
• Picture tests: Write or tell a story in response to a picture
• Thematic apperception test: series of pictures; you tell the story
• Cartoon (balloon) test: fill in the dialogue of a cartoon
• Role playing in a given situation
Analyzing Qualitative Data
•
Inductive Approach
•
Goal is understanding why people do what they do and what
products/service mean to them
•
Insights and theory-development are “bottom up”
•
They emerge as researchers read and interpret responses
•
Insights are “contextualized” within a culture/subculture (thick description)
Nisbett & Wilson (1977, Psych Review)
Telling More Than We Can Know:
Verbal Reports on Mental Processes
1. Not aware of a response (snake phobics)
2. Not aware of a stimulus (the cord puzzle)
3. Not aware of a connection between stimulus and
response (nylons)
4. So what? So this: we may need more creative
interview techniques to get into people’s minds.
5. Enter Dr. Clotaire Rapaille 
Dr. Clotaire Rapaille
Archetype Discoveries Worldwide
http://www.rapailleinstitute.com/
I don’t care what you’re going to tell me intellectually.
I don’t care. Give me the reptilian. Why?
Because the reptilian always wins.
Dr. Clotaire Rapaille
• Internationally known expert in Archetype Discoveries and Creativity
• Archetype: In psychology, according to the theory of psychologist Carl Jung, an idea
or way of thinking that has been inherited from the experience of the race and
remains in the consciousness of the individual, influencing his perception of the
world. (Webster’s)
• Dr. Rapaille's technique for market research based on his work in the areas of
psychiatry, psychology, and cultural anthropology.
• Dr. Rapaille searches for the “code” behind certain words and ideas (e.g., luxury),
and uses these insights to help marketers promote their products.
Dr. Clotaire Rapaille
• On the Limitations of Traditional Marketing Researchers:
• “They are too cortex, which means that they think too much, and then they ask people
to think and to tell them what they think. Now, my experience is that most of the time,
people have no idea why they’re doing what they’re doing. They have no idea, so they’re
going to try to make up something that makes sense. Why do you need a Hummer to go
shopping? “Well, you see, because in case there is a snowstorm.” No. Why [do] you buy
four wheel drive? “Well, you know, in case I need to go off-road.” Well, you live in
Manhattan; why do you need four wheel drive in Manhattan? “Well, you know,
sometime[s] I go out, and I go—” You don’t need to be a rocket scientist to understand
that this is disconnected. This is nothing to do with what the real reason is for people to
do what they do. So there are many limits in traditional market research.”
• Dr. Rapaille in action: Finding the code for “luxury”  (42:20)
The Reptilian Brain
Reptilian
Oldest part of brain
from an evolutionary
perspective
Paul D. MacLean (1913 - 2007)
American physician
Neuroscientist
Yale, NIMH
Triune Brain Theory
• Reptilian brain (instincts)
• Limbic system (emotion)
• Neocortex (higher order thought)
Descriptive Designs:
Surveys & Observations
Chapter 8
Is X related to Y?
When Are Descriptive Designs Appropriate?
1. Want to describe current characteristics of a market (e.g.,
attitudes toward an existing product or certain aspects of
the marketing mix)
2. Want to understand your target market’s characteristics
(e.g., demographics, psychographics)
3. Want to understand relationships between variables (e.g.,
price and purchase) or differences between groups (e.g.,
attitudes toward water filters between hikers and
backpackers)
Sampling vs. Nonsampling Errors
• Sampling Error
• Statistically speaking, the difference between the sample
results and the population parameter
• Assuming perfect survey, sampling frame, execution, and
respondents, we will still have error due to sampling
• Sampling error becomes smaller with larger sample
• Nonsampling (or Systematic) Error
• A variety of errors that are not related to sampling error
and/or sample size
Four Characteristics of Systematic Error
• Nonsampling (Systematic) Error …
1. Leads to “systematic variation” in responses (e.g., skewed
toward more socially desirable responses)
2. Is controllable (e.g., via good survey design and
procedures)
3. Can not be estimated (whereas sampling error can be
estimated; margin of error in a poll)
4. Are interdependent (i.e., one type of systematic error can
lead to another)
•
Conceptual breakdown (Exhibit 7.2) – Handout in class
Non-Response Errors
• Non-response error occurs when…
• The final sample differs from the planned sample
• Often happens when you can’t contact those in the planned sample or
they refuse to participate
• Those who choose not to respond often of lower income, education, and
more likely to be male
• Non-response can limit generalizability of findings to broader population
• Strategies for reducing non-response error
– Create good rapport, respect respondent’s time, enhance credibility of
research sponsor, use shorter questionnaires
Response Error (Bias)
• Response error occurs when…
• The responses people give are not accurate
• May occur due to
• Deliberate falsification (e.g., social desirability, hostility)
• Unconscious misrepresentation (e.g., faulty memory, desire to please
researcher)
• Might be able to detect with reaction times
• Very fast or very slow RTs may tell you something
Sampling Errors
• Population specification (frame) error
• Your population is all Republicans, but you define your population as
Republicans in WA
• Sample selection error
• When an inappropriate sample is selected from the desired population
• May be due to either poor sampling procedures or intentionally
excluding certain people from the sample
• Sample frame error
• Sample frame = list of potential people in your target population
• Sample frame error = when the sample frame is not representative of
your population (e.g., only those with email addresses)
Four Broad Categories of Survey Methods
• Person Administered
• In-home, executive, mall-intercept, purchase-intercept
• Telephone Administered
• Either by a person or completely automated
• Self Administered
• Panels, drop off, mailed survey
• Computer Assisted
• Fax, email, internet
Person Administered Surveys
Advantages
Disadvantages
•
Interviewer can adapt to respondent
•
Can be slow
•
Interviewer can create good rapport
•
Interviewers may incorrectly interpret
response (selective listening)
with respondents
•
Interviewer can clarify questions and
•
“correct” response
get insight via non-verbal responses
•
Interviewers can ensure they are
sampling the correct people
Interviewers may give off “clues” to the
•
Can be expensive
Telephone Surveys
Advantages
Disadvantages
•
•
Can monitor interviewers for quality
control
•
Can’t use visual stimuli (though might
be possible with cell phones)
Less expensive than person
administered
•
Can be hard to keep a large amount of
info in memory during interview
•
Following up if respondent not available
first time is inexpensive
•
•
People bail on long phone interviews
•
Public is distrusting; can limit sample
People who don’t agree to person
administered (e.g., due to time
constraints) may be more willing to do a
telephone interview
Self Administered Surveys
Advantages
Disadvantages
•
•
Low cost (no need for interviewer)
Can’t obtain any information beyond
what is presented on survey (no follow
up questions or probing possible)
•
Respondents not rushed, can take
time if they want to
•
Interviewer can’t bias response
•
Low response rates
•
If respondent doesn’t understand, can’t
ask an interviewer; may lead to response
errors
•
Anonymity can lead to more truthful
responses
•
Data comes in slowly; may require several
re-contacts
Paco Underhill
The “King”
of Observation
www.envirosell.com
When to Use Observation
• When the respondent may not be able to accurately recall the frequency
of a behavior, and/or may be inclined to give misleading answers
• When the response in question is a behavior (rather than a feeling)
• When the behavior in question is relatively frequent and occurs within a
limited time frame
• When the behavior in question can be observed (e.g., in public)
Observation
Advantages
Disadvantages
•
•
Gain data on actual behavior (rather
Generalizing from a limited number of
observations can be difficult
than self-reported behavior which
may be biased)
•
May be difficult to understand why the
behavior occurred
•
If doing observation in person (not
recorded), possible to miss important
behaviors (or other people)
Causal Designs
Chapter 9
Understanding when (and why)
XY
Theories and Hypotheses
•
Theory
•
•
A body of interconnected propositions about how a phenomenon
works (recall animosity model)
Hypothesis Testing
•
Null (Dull) Hypothesis (Ho):
• Nothing interesting is going on
• Any differences we are observing are completely due to chance
•
Alternative Hypothesis (H1)
• Something interesting is going on
• Differences in DV are due to IV
Experiments: Some Key Definitions
•
Independent Variable (X, the cause, the predictor)
•
The variable you “manipulate” (good vs. bad aroma in store)
•
Dependent Variable (Y, the outcome, the criterion)
•
What happens after you manipulate the IV (sales of a product)
•
Control variables
•
Variables that you don’t allow to vary along with the IV
•
If any variable covaries with the IV, then there is a confound (e.g., if
music systematically varies along with aroma, then you can’t tell if it’s
the aroma or the music that influences sales)
•
Extraneous variables (or noise)
•
“Stuff happens” during an experiment, but it evens out across the levels
of the independent variable (e.g., different music at different times, but
it doesn’t systematically vary by the level of the IV)
Factorial Designs
•
Factorial Design
•
When the researcher is examining the impact of two independent (or
predictor) variables on a DV
•
Can have two main effects (overall impact of each IV) and an interaction
(combined effect of two IVs)
Low Hostility
High Hostility
Column Means
No Service
Failure
2
2
Service
Failure
3
5
2
4
Main Effect of
Service Failure
Row Means
2.5
3.5
Main Effect of
Hostility
Interaction 
(Moderation)
Moderation
Moderation (under what conditions is a relationship stronger/weaker)
•
When the effect of one IV (service failure) on the DV (negative word of
mouth) depends on the level of another IV (trait hostility)
Service Failure  NWOM
Trait
Hostility
Negative Word of Mouth
(DV)
•
6
Trait Hostility
(IV 2)
5
Low
Hostility
4
3
High
Hostility
2
1
0
No Failure
Service Failure
Service Failure (IV 1)
Mediation
•
Mediation (like a combination shot in pool)
•
When the effect of one IV on the DV occurs through an “intermediary”
variable (think cue ball hits one ball hits eight ball)
•
For example, assume a person experiences a service failure
•
They infer a negative motive, feel angry, and spread NWOM
•
Here, anger is the mediator between Inference of Negative Motive and
NWOM
Inference of
Negative Motive
Anger
Negative Word
Of Mouth
Validity: Some Key Definitions
•
Validity (in general)
•
The extent to which conclusions drawn from a study are true
•
Internal Validity
•
When a researcher can clearly identify cause and effect relationships (i.e.,
there are no confounds)
•
External Validity
•
The extent to which what you find in your study can be generalized to your
target population
•
Construct Validity
•
Extent to which your constructs of interest (e.g., sensation seeking) are
accurately and completely identified (measured)
•
In other words, the extent to which you are actually measuring what you say
you are measuring (your sensation seeking scale really does measure the
true construct of sensation seeking)
Threats to Internal Validity
•
•
•
•
•
•
•
History Effect
•
When something (an historical event) happens during the course of a study that
affects the dependent variable
Maturation Effect
•
Similar to a history effect; something happens over time (changes in the individual)
that affects the DV
Testing Effect
•
In a pretest-posttest design, you affect the time 2 DV by pretesting at time 1; the
simple act of measuring the DV at time 1 changes the DV at time 2
Instrumentation Effect
•
The mere fact that you are measuring something (e.g., observing behavior) changes
the behavior
Statistical Regression
•
When you select groups based on extreme scores, they regress toward the mean,
changing your groups
Selection Bias
•
When groups (control, experimental) differ before experimental manipulation; creates
unequal groups (a confound)
Mortality
•
Some drop out or die (attrition), and these drop-outs change scores in the condition;
those who stick around may be different than those who drop out
Experimental Research Designs
•
Terminology
•
X = subjects are exposed to a treatment (independent variable)
•
O = the outcome (dependent variable)
•
[R] = random assignment of subjects to conditions
•
EG = experimental group
•
CG = control group
•
 = time
•
Pre-experimental (“Crude” experimental) designs
•
Either have no control group or non-random assignment to groups
•
Suffer from low internal validity because it is not possible to compare
groups without the possibility of confounding factors
•
Types: One-shot, One-group, and Static Group comparisons
Pre-Experimental Research Designs - 1
•
One Shot Study
•
X  O1
•
How do customers respond (O) to a single product like gatorade (X)?
•
Problem? No control group. Response could be driven by many factors
that covary along with the product (e.g., lighting, context). That is, there
are many opportunities for extraneous variables to “confound” the
manipulation of the IV.
•
Internal validity low
Pre-Experimental Research Designs - 2
•
One Group Pre-test Post-Test
•
O1  X  O2
•
How do sales of sweaters at time 1 (O1) change at time 2 (to O2) after the
introduction of a new product display (X)?
•
Problem? No control group. History and Testing Effects. Response could
again be driven by many factors that covary along with the manipulation
of the product display, that change over time with the introduction of
the new product display (e.g., changes in store music or changes in the
economy), or are related to testing at time 1. In other words, there are
many opportunities for extraneous variables to “confound” the
manipulation of the IV.
•
Internal validity low
Pre-Experimental Research Designs - 3
•
Static Group Comparison (Non-random assignment to groups)
•
•
Experimental Group (EG): X  O1
Control Group (CG):
 O2
•
Compare two stores. In Store 1 (EG), use a promo display for nose strips.
In Store 2 (CG), don’t use a promo display. Compare sales.
•
Problem? Non-random assignment to groups. This again allows factors
other than the promo display to affect sales. For example, in Store 1, it
could be that the pharmacists are more friendly and more likely to
recommend nose strips to their weary-eyed customers.
•
Internal validity low
True Experimental Research Designs - 1
•
•
Pre-test, Post-test Control Group (subjects randomly assigned to groups)
Also called a mixed design (one within-subject variable, time; and one
between-subject variable, the experimental manipulation)
•
•
•
Experimental Group (EG): [R] O1  X  O2
Control Group (CG):
[R] O3 …….. O4
Treatment effect = (O2-O1) – (O4-O3): The difference between differences
•
Test all subjects (O1, O3), then randomly assign to experimental or
control group, then test again (O2, O4)
•
Eliminates testing effects, maturation, and (with good control over
experimental conditions) confounding factors.
•
Internal validity higher than earlier designs, but if not careful (low
control over conditions), internal validity could be threatened
True Experimental Research Designs - 2
•
Post-test Only (but with subjects randomly assigned to groups)
•
•
Experimental Group (EG): [R] X  O1
Control Group (CG):
[R]
O2
•
Randomly assign to experimental group or control group, then compare
levels of dependent variable (O)
•
No testing effects. With good control over experimental conditions,
eliminates confounding factors.
•
Internal validity higher than static group comparison, but if not careful
(low control over conditions), internal validity could be threatened.
True Experimental Research Designs - 3
•
Solomon Four Group Design
•
Design 1 (Pre-test, post-test control group design)
•
•
•
Experimental Group (EG):
Control Group (CG):
[R] O1  X  O2
[R] O3 …….. O4
Design 2 (Post-test only design)
•
•
Experimental Group (EG):
Control Group (CG):
•
Let’s say that O = anger with waiting in line, and X = pleasant fragrances.
•
If [O2 < O1], [O2 < O4], [O5 < O6], [O5 < O3], strong internal validity!
[R]
[R]
X O5
O6
Quasi-Experimental Research Designs - 1
•
•
Non-equivalent control group
Like pre-test, post-test control group, but it is groups of subjects (not
individual subjects) who are randomly assigned to conditions. Hence there is
no [R] shown below. For example, you could randomly assign stores to
conditions (experimental vs. control), but you can’t randomly assign people
to conditions, and you can’t control everything about the stores that may be
confounded with the experimental manipulation.
•
•
•
•
•
Experimental Group (EG): O1  X  O2
Control Group (CG):
O3 …….. O4
Treatment effect = (O2-O1) – (O4-O3): The difference between differences
Equivalence of groups prior to treatment: (O3-O1)
Eliminates testing effects, maturation, and (with good control over
experimental conditions) confounding factors. Scores at pre-test (O1 and O3)
can be used as a control variable in data analysis. If in field, external validity
is heightened over straight lab studies.
Quasi-Experimental Research Designs - 2
•
•
•
•
•
Separate sample pretest-posttest
Some folks (Sample 1) are tested before an advertising campaign (O1)
Then an advertising campaign occurs (X)
Then another group is tested after the campaign (O2)
Can’t be sure people were exposed to treatment (X), which is why it’s in
parentheses
•
•
•
Sample 1: O1  (X)
Sample 2:
(X)  O2
Some problems with internal validity (history, maturation), but external
validity is high due to its naturalistic setting
Sampling
It is often said that without water,
life would be impossible.
Similarly, without sampling, marketing
research as we know it would be impossible.
Feinberg, Kinnear, & Taylor (2008, p. 290)
Probability vs. Nonprobability Sampling
• Probability Sampling
• Each sampling unit has a known probability of being
included in the sample
• Nonprobability sampling
• When the probability of selecting each sampling unit is
unknown
Probability Sampling Procedures
• Simple Random Sampling
• A sampling approach in which each sampling unit in a target
population has a known and equal probability of being included
• Advantage: Good generalizability and unbiased estimates
• Disadvantage: must be able to identify all sampling units within
a given population; often, this is not feasible
• Systematic Random Sampling
• Similar to random sampling, but work with a list of sampling
units that is ordered in some way (e.g., alphabetically).
• Select a starting point at random, then survey each nth person
where the “skip interval” = (population size/desired sample size)
• Advantage: quicker and easier than SRS
• Disadvantage: may be hidden “patterns” in the data
Probability Sampling Procedures
• Stratified Random Sampling
• Break up population into meaningful groups (e.g., men, women),
then sample within each “strata”, then combine
• Proportionate stratified sampling: here you sample based on the
size of the populations (i.e., sample more from the bigger strata:
e.g., Caucasians)
• Disproportionate stratified sampling: sample the same number of
units from each strata, regardless of the strata’s size in the pop.
• A variant is optimal allocation: here you use smaller sample
sizes for strata within which there is low variability (as the lower
variability will give you more precision with lower N).
• Advantages: more representative; can compare strata
• Disadvantages: Can be hard to figure out what to base strata on
(Gender? Ethnicity? Political party?)
Probability Sampling Procedures
• Cluster Sampling
• Similar to stratified random sampling, but with stratified random
sampling, the strata are thought to possibly differ between strata
(men vs. women), but be homogeneous within strata.
• In cluster sampling, you divide overall population into
subpopulations (like SRS), but each of those subpopulations (called
“clusters”) are assumed to be mini-representations of the
population (e.g., survey customers at 10 Red Robins in WA).
• Area sampling: clusters based on geographic region
Probability Sampling Procedures
• Cluster Sampling
• One-step clustering: just select one cluster (e.g., one store);
problem = may not be representative of population
• Two-step cluster sampling: break into meaningful subgroups (Red
Robins in big cities vs. Red Robins in suburbs), then randomly
sample within each of those clusters
• Advantages: easy to generate sampling frame; cost efficient;
representative; can compare clusters
• Disadvantages: must be careful in selecting the basis for clusters;
also, within clusters, often little variability (they’re homogeneous),
and this lack of variability leads to less precise estimates
Nonprobability Sampling Procedures
•
•
•
•
Convenience Sample
Survey people based on convenience (e.g., college students)
Advantage: is fast and easy
Disadvantage: may not be representative
•
•
•
•
Judgment Sampling
Use your judgment about who is best to survey
Advantage: Can be better than convenience if judgment is right
Disadvantage: but if judgment wrong, may not be
representative/generalizable
Nonprobability Sampling Procedures
• Quota Sampling
• Sample fixed number of people from each of X categories, possibly
based on their relative prevalence in the population
• Advantage: Can ensure that certain groups are included
• Disadvantage: but b/c you aren’t using random sampling,
generalizability may be questionable
• Snowball Sampling
• You contact one person, they contact a friend (e.g., one cancer
survivor is in contact with other survivors, and so recruits them)
• Advantages: can make it easier to contact people in hard to reach
groups
• Disadvantage: there may be bias in the way people recruit others
Factors Affecting Choice of Sampling Procedure
• Use some type of random sampling if:
• You are collecting quantitative data that you want to use to arrive at
accurate generalizations about population
• You have sufficient resources and time
• You have a good sense for the population
• You are sampling over a broader range (e.g., of states, nations)
Computing the Sample Size Based on Usable Rates
• Several factors can reduce your sample size
• Thus, you may want to plan for more than your final sample size
(i.e., use a higher “number of contacts” to achieve your final sample
size). You adjust using the following three factors:
• RR = reachable rate (e.g., how many people on a telephone list
will you actually be able to reach?)
• OIR = overall incidence rate (i.e., % of target population that will
qualify for inclusion; e.g., can’t use people over 40)
• ECR = expected completion rate (i.e., some folks won’t complete
your survey)
• For example, -
Computing the Sample Size Based on Usable Rates
• You want a sample size of n = 500
• You figure you can reach 95% of the folks on your list (RR = .95)
• You think 60% will be 40 or younger (OIR = .60)
• You predict that 70% will complete your survey (ECR = .70)
• Based on these numbers, you should contact 1,253 people
Number of contacts
n
500

 1,253
(RR) x (OIR) x (ECR) (.95)(.60)(.70)
Some Key Terms
•
•
•
•
•
•
Sampling
• Selection of a small number of elements from a larger defined target group of
elements and expecting that the information gathered from the small group
will allow judgments to be made about the larger group
Population
• The identifiable set of elements of interest to the researcher and pertinent to
the information problem
Defined Target Population
• The complete set of elements identified for investigation
Element
• A person or object (e.g., a firm) from the defined target population from
which information is sought
Sampling Units
• The target population elements available for selection during the sampling
process
Sampling Frame
• The list of all eligible sampling units
Some Key Terms
• Total Error = Sampling Error + Nonsampling Error
• Sampling Error
• Any type of bias that is attributable to mistakes in either drawing a
sample or determining the sample size
• Nonsampling Error (controllable)
• A bias that occurs in a reesearch study regardless of whether a sample
or a census is used (recall all the different types of errors we
discussed)
• Respondent Errors (non response, response errors)
• Researcher’s measurement/design errors (survey, data analysis)
• Problem definition errors
• Administrative errors (data input errors, interview errors, poor
sample design)
Central Limit Theorem
• A theory that states that, regardless of the shape of the
population from which we sample (e.g., positively skewed), as
long as our sample size is > 30, the sampling distribution of
the mean (x-bar) will be normally distributed with the
following characteristics:
x
sx 
The mean of the sampling distribution of the mean
will equal the mean in the population.
s
n
The standard error of the sampling distribution of
the mean will equal sample standard deviation (s)
divided by sample size (n). This is a sample
estimate of the true standard error in population.
The larger the sample size, the more precise we can
get about our estimate of the true mean in the
population (e.g., in our confidence interval).
variance
Note: Dr. Joireman does not put
a “bar” above s or s2.
Computing Standard Deviation
Assume your data are continuous
(i.e., are not just yes/no data).
For example, let’s say we want to know how much
people would be willing to pay for a tennis racquet.
We sample 7 folks and wish to generalize to the
population….
Results 
Formulas for
Variance and Standard Deviation
Sum of Squared
Deviations
POPULATION
SS
Population Variance   
N
2
PopulationStandard Deviation   
SS
N
SAMPLE
Sample Variance  s 2 
SS
N 1
SampleStandard Deviation  s 
SS
N 1
The Sum of Squared Deviations (SS)
Concept ualFormula
SS  ( X i  X ) 2
Highlights Concept
Tells a Story
R aw S core F ormula
SS  X
2
( X ) 2

N
“Crank it Out”
Faster, Less Meaningful
• Both Formulas Give Identical Answers!
• SS = NUMERATOR of the Variance
• Examples on board…
Example of Computing Standard Deviation (for a Sample)
Xi
Mean
Xi-Mean
(Xi-Mean)2
X2
60
75
-15
225
3600
65
75
-10
100
4225
70
75
-5
25
4900
75
75
0
0
5625
80
75
5
25
6400
85
75
10
100
7225
90
75
-15
225
8100
Σ(X-M) = 0 !
Σ(X-M)2 = 700
ΣX2 = 40075
ΣX = 525
Concept ualFormulafor SS
SS  ( X i  X ) 2  700
Raw Score Formulafor SS
 (X ) 2 
 (525) 2 
SS  X  
  40075 
  40074 39375 700
 N 
 7 
2
Sample Variance  s 2 
SS
700

 116 .67
N 1 7 1
Sample Standard Deviation s 
Standard Errorof Mean  s X 
SS
700

 10.82
N 1
7 1
s
10.82

 4.09
n
7
This is the “standard deviation” of the sampling distribution of means.
This (4.09) will naturally be smaller than our sample standard deviation (10.82)
based on our single sample of scores, and it will become smaller as n increases.
Confidence Intervals
A confidence interval is the statistical range of
values within which the true value of the
target population parameter is expected to lie.
Common Z-Critical Values
• To be 90% confident, you use a z-critical value of 1.65
• To be 95% confident, you use a z-critical value of 1.96
• To be 99% confident, you use a z-critical value of 2.58
An example…
Z-critical values for 95% confidence
(put ½ of .05 on each side)
.025
.025
-1.96
+1.96
Computing Confidence Intervals
• 95% Confidence Interval:
• We are 95% confident that the mean of the population from which we
took our sample has a mean between these lower and upper limits.
• To compute, we need:
Mean of our sample
Standard error
of mean
Critical Z-value for our
desired level of confidence
(see next page for Z-critical values)
ConfidenceInterval.95  CI  x  (sx )(Zb, cl)  75  (4.09)(1.96)  75  8.02
Restated,ConfidenceInterval.95  66.98 X  83.02
Based on these results, we are 95% confident that the mean in the population
from which we sampled is between 66.98 and 83.02. Cool beans!

similar documents