Methodology in Marketing and
Strategy Research Lectures
Lecture 1
Introduction lecture of the course
Assignments will in total yield 0.2 bonus points if done above average
Second part of the lecture
Construct
- Definition: a conceptual term used to describe a phenomenon of theoretical interest. All
constructs should have a clear definition
- A construct is quantifiable and directly/indirectly observable.
- An indirectly observable construct is called latent
- Examples from strategy research: firm success, competitive advantage, innovation
performance
Construct must be defined in terms of
- Object – something that is looked at
- Attribute – it needs to describe something
- Rater entity – the level on which you look at the construct – e.g. in the example of team
diversity, the rater entity is team.
Overview of relationships
- Direct causal relationships
- (fully or partially) mediated relationships (indirect) causal relationships
- Spurious relationships
- Bidirectional (cyclic) causal relationships
Direct causal relationships
- A leads to B. In social sciences, this usually implies a linear effect.
Mediated causal relationship
- A appears to have a direct effect on B. But they are not directly related but through a
mediator Z. A influences Z, Z influences B.
- Full mediation: effect of A on B is completely mediated by Z
- Partial mediation: A influences Z, Z influences B, but A still influences B directly (for a smaller
part)
Exogenous variable: a variable whose value is determined outside the model and is imposed on the
model.
,Endogenous variable: a variable whose value is determined by the model. The moment there is an
arrow on a variable, it is an endogenous variable.
Spurious relationship
- A third variable influences A as well as B
- Example: there is a relationship between drowning in pools and ice cream consumption.
Does ice cream then lead to drowning in pools? No, both are influenced by warm weather.
Bidirectional causal relationship
- A leads to B, and B leads to A. There is a feedback loop.
- But, this does not necessarily happen at the same time.
- Not so common in social science research because it’s hard to prove and difficult to entangle.
- Chicken and egg problem
Unanalysed relationship
- There is a correlation between A and B which is not analysed further.
Moderated causal relationship
- The strength and/or direction of the effect of A on B differs depending on the level of M.
- Variable M is a moderator
Theory of reasoned action
- Explains why we do something
- Norms and attitude lead to a certain intention, which in turn can lead to an action.
Greek letters: theory level
X and Y: observational level
Measurement model: first step, measuring all the constructs. Corresponding to the observation
Structural model: corresponding to the theoretical model, measuring the relationships between the
constructs.
Multi-item measurement
- In the example the different constructs have several items
- Using several items increases reliability and validiy of measures.
- It also allows for measurement assessment: measurement error, reliability, validity.
- Two forms of measurement models:
Formative (emerging)
Reflective (latent)
,Reflective measurement models:
- Direction of causality is from the construct to the measure
- The observation drives the measurement
- The indicators are expected to be correlated
- Dropping an indicator from the measurement model does not alter the meaning of the
construct
- Similar to factor analysis
Formative measurement models:
- Direction of causality is from measure to construct
- No reason to expect indicators to be correlated
- Dropping an indicator from the measurement model may alter the meaning of the construct
- Based on multiple regression
- Important to have all possibilities in your measurement, e.g. in indexes. Think about the
drunkenness example: if you only measure wine & liquor consumption and someone only
drank beer, you may think someone is not drunk.
Technique Consultation Factor Analysis
, Principal components analysis
- Takes into account the total variance
- Example, you have 5 variables. Then PCA looks into the total variance of these 5 variables.
- If your goal is to explain as much variance as possible, you would use PCA. It is more
exploratory way.
- Primary concern: minimum number of factors that account for the maximum of variance.
- Factors are called principal components.
- Mathematically, each variable is expressed as a linear combinations of the components. The
covariation among the variables is described in terms of a small number of principal
components.
- PCA standardizes the variables – standardized regression coefficient.
Common factor analysis
- Factors are estimated based only on the common variance. So not the total amount of
variance is taken into account. We speak of communalities.
- Common item variance is the variance one item shares with all other items included.
- Primary concern: identify underlying dimensions and their common variance.
- Mostly used in research in which there is also prior knowledge, so that could be in a thesis.
- In SPSS this is called principal axis factoring.
SPSS Example – Principal components analysis
- All variance has been extracted. You see communalities all of 1.
SPSS Example – Common factor analysis
- Extraction is the variance shared by variables. You see these are not only 1.
Initial communalities – based on maximum number of factors possible, extracted communalities are
based on the numbers of factors that can be extracted (e.g. factors with an eigenvalue of > 1)
Factor loading: the correlation between factor and item.
Main differences between PCA and CFA
- Amount of variance they take along
- Different purposes. PCA is exploratory, CFA is more confirmatory in that there is prior
knowledge.
Factor rotation
- Why do we rotate at all?
We move the factors closer to the real observations, in order for the results to be more
interpretable.
- Orthogonal rotation is the default rotation.
Strategy Research Lectures
Lecture 1
Introduction lecture of the course
Assignments will in total yield 0.2 bonus points if done above average
Second part of the lecture
Construct
- Definition: a conceptual term used to describe a phenomenon of theoretical interest. All
constructs should have a clear definition
- A construct is quantifiable and directly/indirectly observable.
- An indirectly observable construct is called latent
- Examples from strategy research: firm success, competitive advantage, innovation
performance
Construct must be defined in terms of
- Object – something that is looked at
- Attribute – it needs to describe something
- Rater entity – the level on which you look at the construct – e.g. in the example of team
diversity, the rater entity is team.
Overview of relationships
- Direct causal relationships
- (fully or partially) mediated relationships (indirect) causal relationships
- Spurious relationships
- Bidirectional (cyclic) causal relationships
Direct causal relationships
- A leads to B. In social sciences, this usually implies a linear effect.
Mediated causal relationship
- A appears to have a direct effect on B. But they are not directly related but through a
mediator Z. A influences Z, Z influences B.
- Full mediation: effect of A on B is completely mediated by Z
- Partial mediation: A influences Z, Z influences B, but A still influences B directly (for a smaller
part)
Exogenous variable: a variable whose value is determined outside the model and is imposed on the
model.
,Endogenous variable: a variable whose value is determined by the model. The moment there is an
arrow on a variable, it is an endogenous variable.
Spurious relationship
- A third variable influences A as well as B
- Example: there is a relationship between drowning in pools and ice cream consumption.
Does ice cream then lead to drowning in pools? No, both are influenced by warm weather.
Bidirectional causal relationship
- A leads to B, and B leads to A. There is a feedback loop.
- But, this does not necessarily happen at the same time.
- Not so common in social science research because it’s hard to prove and difficult to entangle.
- Chicken and egg problem
Unanalysed relationship
- There is a correlation between A and B which is not analysed further.
Moderated causal relationship
- The strength and/or direction of the effect of A on B differs depending on the level of M.
- Variable M is a moderator
Theory of reasoned action
- Explains why we do something
- Norms and attitude lead to a certain intention, which in turn can lead to an action.
Greek letters: theory level
X and Y: observational level
Measurement model: first step, measuring all the constructs. Corresponding to the observation
Structural model: corresponding to the theoretical model, measuring the relationships between the
constructs.
Multi-item measurement
- In the example the different constructs have several items
- Using several items increases reliability and validiy of measures.
- It also allows for measurement assessment: measurement error, reliability, validity.
- Two forms of measurement models:
Formative (emerging)
Reflective (latent)
,Reflective measurement models:
- Direction of causality is from the construct to the measure
- The observation drives the measurement
- The indicators are expected to be correlated
- Dropping an indicator from the measurement model does not alter the meaning of the
construct
- Similar to factor analysis
Formative measurement models:
- Direction of causality is from measure to construct
- No reason to expect indicators to be correlated
- Dropping an indicator from the measurement model may alter the meaning of the construct
- Based on multiple regression
- Important to have all possibilities in your measurement, e.g. in indexes. Think about the
drunkenness example: if you only measure wine & liquor consumption and someone only
drank beer, you may think someone is not drunk.
Technique Consultation Factor Analysis
, Principal components analysis
- Takes into account the total variance
- Example, you have 5 variables. Then PCA looks into the total variance of these 5 variables.
- If your goal is to explain as much variance as possible, you would use PCA. It is more
exploratory way.
- Primary concern: minimum number of factors that account for the maximum of variance.
- Factors are called principal components.
- Mathematically, each variable is expressed as a linear combinations of the components. The
covariation among the variables is described in terms of a small number of principal
components.
- PCA standardizes the variables – standardized regression coefficient.
Common factor analysis
- Factors are estimated based only on the common variance. So not the total amount of
variance is taken into account. We speak of communalities.
- Common item variance is the variance one item shares with all other items included.
- Primary concern: identify underlying dimensions and their common variance.
- Mostly used in research in which there is also prior knowledge, so that could be in a thesis.
- In SPSS this is called principal axis factoring.
SPSS Example – Principal components analysis
- All variance has been extracted. You see communalities all of 1.
SPSS Example – Common factor analysis
- Extraction is the variance shared by variables. You see these are not only 1.
Initial communalities – based on maximum number of factors possible, extracted communalities are
based on the numbers of factors that can be extracted (e.g. factors with an eigenvalue of > 1)
Factor loading: the correlation between factor and item.
Main differences between PCA and CFA
- Amount of variance they take along
- Different purposes. PCA is exploratory, CFA is more confirmatory in that there is prior
knowledge.
Factor rotation
- Why do we rotate at all?
We move the factors closer to the real observations, in order for the results to be more
interpretable.
- Orthogonal rotation is the default rotation.
