Marketing Research Methods
Table of Contents
Standard information................................................................2
Principal component analysis....................................................4
Exploratory factor analysis........................................................7
Reliability analysis....................................................................9
ANOVA & ANCOVA...................................................................10
ANOVA assumptions................................................................................................. 11
2-way ANOVA without interaction.............................................................................12
2-way ANOVA with interaction..................................................................................12
Regression.............................................................................. 15
Bivariate regression.......................................................................15
Assumptions of regession.........................................................................................16
Dummy coding..............................................................................20
Binary regression....................................................................20
Linear probability model (LPM).......................................................21
Logistic regression (logit)..............................................................24
Model fit: how good is our model?..................................................25
Naïve model: benchmark model......................................................26
Moderation analysis................................................................27
Rescaling of variables: mean-centering...........................................28
Rescaling of variables: log-transformation......................................29
Moderator variables.......................................................................30
Single dummy moderator.........................................................................................30
Multiple dummy moderators.....................................................................................30
Single continuous moderator....................................................................................32
Output of moderated regression.....................................................33
Mediation analysis..................................................................34
Interpretation of mediation............................................................38
Other types of mediation:..............................................................39
Cluster analysis......................................................................40
Hierarchal clustering.....................................................................41
Testing cluster solutions................................................................45
Validating cluster solutions............................................................47
,Standard information
Conceptualization = drawing boundaries around terms to make them
tangible.
,Research design types
- Exploratory: when you have no idea what you want to do yet.
- Discover ideas, insights, understanding processes
- Experts, qualitative data
- Descriptive: how these relations relate to each other
- Describing important characteristics/markets
- Surveys, panels, quantitative data
- Causal:
- Determine cause-effect relations
Cleanest experiments
Experiments
- Manipulation of one or more independent variables
- Measurement of the effect on one or more dependent variables.
- You can make statements about causality because of experimental
control: nothing except the independent variable is changing.
Within-subjects design = each participant provides data for all the levels
of the independent variables; aka repeated measures design.
- Advantages: holds subject variables constant, increase statistical
power by reducing random variation, reduces the number of
subjects needed.
- Disadvantages: potential threats to validity maturation,
instrumentation, testing.
Between-subjects design = each participant receives only one level of the
independent variable.
- Advantages: avoids between-experiments comparison
- Disadvantages: large designs require large number of participants
(at least 75), must address selection issue (random assignment),
lack of statistical power.
, Always do a manipulation check. Is my variable well-perceived. Is there a
difference between high and low. Also do an attention check.
Validity in experiments
- Internal validity: did the independent variable actually cause the
effect on the dependent variable. No effect of extraneous variables?
- External validity: can the results of the experiment be generalized in
terms of; population, geographic areas, product categories.
- Higher internal validity often means lower external validity. But field
experiments!
Experimental designs: three main classes
- True experiments
- Lab, online, fully controlled
- Allows for measuring the underlying process
- Field experiments
- Random assignment, observations
- Does not allow for measuring the underlying process
- Quasi experiments
- No random assignment
Principal component analysis
Goal: Data reduction technique.
Transforms a large set of correlated variables into a smaller set of
uncorrelated components that still capture most of the variance in data.
Table of Contents
Standard information................................................................2
Principal component analysis....................................................4
Exploratory factor analysis........................................................7
Reliability analysis....................................................................9
ANOVA & ANCOVA...................................................................10
ANOVA assumptions................................................................................................. 11
2-way ANOVA without interaction.............................................................................12
2-way ANOVA with interaction..................................................................................12
Regression.............................................................................. 15
Bivariate regression.......................................................................15
Assumptions of regession.........................................................................................16
Dummy coding..............................................................................20
Binary regression....................................................................20
Linear probability model (LPM).......................................................21
Logistic regression (logit)..............................................................24
Model fit: how good is our model?..................................................25
Naïve model: benchmark model......................................................26
Moderation analysis................................................................27
Rescaling of variables: mean-centering...........................................28
Rescaling of variables: log-transformation......................................29
Moderator variables.......................................................................30
Single dummy moderator.........................................................................................30
Multiple dummy moderators.....................................................................................30
Single continuous moderator....................................................................................32
Output of moderated regression.....................................................33
Mediation analysis..................................................................34
Interpretation of mediation............................................................38
Other types of mediation:..............................................................39
Cluster analysis......................................................................40
Hierarchal clustering.....................................................................41
Testing cluster solutions................................................................45
Validating cluster solutions............................................................47
,Standard information
Conceptualization = drawing boundaries around terms to make them
tangible.
,Research design types
- Exploratory: when you have no idea what you want to do yet.
- Discover ideas, insights, understanding processes
- Experts, qualitative data
- Descriptive: how these relations relate to each other
- Describing important characteristics/markets
- Surveys, panels, quantitative data
- Causal:
- Determine cause-effect relations
Cleanest experiments
Experiments
- Manipulation of one or more independent variables
- Measurement of the effect on one or more dependent variables.
- You can make statements about causality because of experimental
control: nothing except the independent variable is changing.
Within-subjects design = each participant provides data for all the levels
of the independent variables; aka repeated measures design.
- Advantages: holds subject variables constant, increase statistical
power by reducing random variation, reduces the number of
subjects needed.
- Disadvantages: potential threats to validity maturation,
instrumentation, testing.
Between-subjects design = each participant receives only one level of the
independent variable.
- Advantages: avoids between-experiments comparison
- Disadvantages: large designs require large number of participants
(at least 75), must address selection issue (random assignment),
lack of statistical power.
, Always do a manipulation check. Is my variable well-perceived. Is there a
difference between high and low. Also do an attention check.
Validity in experiments
- Internal validity: did the independent variable actually cause the
effect on the dependent variable. No effect of extraneous variables?
- External validity: can the results of the experiment be generalized in
terms of; population, geographic areas, product categories.
- Higher internal validity often means lower external validity. But field
experiments!
Experimental designs: three main classes
- True experiments
- Lab, online, fully controlled
- Allows for measuring the underlying process
- Field experiments
- Random assignment, observations
- Does not allow for measuring the underlying process
- Quasi experiments
- No random assignment
Principal component analysis
Goal: Data reduction technique.
Transforms a large set of correlated variables into a smaller set of
uncorrelated components that still capture most of the variance in data.
