SUMMARY
INTRODUCTION TO
RESEARCH IN
MARKETING
Demi van de Pol || Master Marketing Analytics & Marketing Management || Tilburg University || 2022
, Demi van de Pol | Summary | Introduction to Research in Marketing | TISEM | Tilburg University | 2022
CONTENT
This summary is written for the course “Introduction to Research in Marketing” during the semester
Spring-2022 and is part of both the master Marketing Analytics and the master Marketing
Management. The input for this summary consists of all the information given in the lectures.
LECTURE 1: Basics for multivariate analyses
DEFINITION MULTIVARIATE ANALYSIS
The term multivariate analysis refers to all statistical methods that simultaneously analyze multiple
measurements on each individual or object under investigation.
BASIC CONCEPTS FOR MULTIVARIATE ANALYSIS
MEASUREMENT SCALES
Measurement scales refer to how different variables can be collected / measured. There can be
made a distinction between non-metric scales (nominal and ordinal) and metric scales (interval and
ratio).
Non-metric measurement scale: Nominal
Nominal scaled variables are used to uniquely identify an object or person or to classify them. In
other words, the goal of nominal scaled variables is unique identification / definition or classification.
Appropriate methods of analysis / statistics for nominal scaled variables can be:
● Percentages (%)
● Mode
● Chi square tests
Non-metric measurement scale: Ordinal
Ordinal scaled variables are used to indicate ‘order’ / sequence. Examples of ordinal scaled variables
can be ranking, level of education, etc.
Appropriate methods of analysis/statistics for ordinal scales can be:
● Percentiles
● Median
● Rank correlation
+ all previous statistics of the nominal scale
Metric measurement scale: Interval
One important characteristic of an interval scaled variable is that the origin is arbitrary, meaning that
the origin not always starts at zero. Interval scaled variables allow for calculations.
Appropriate methods of analysis/statistics for interval scales can be:
● Arithmetic average
● Range
● Standard deviation
● Product-moment correlation
+ all previous statistics of the nominal scale and ordinal scale
1
, Demi van de Pol | Summary | Introduction to Research in Marketing | TISEM | Tilburg University | 2022
Metric measurement scale: Ratio
An important characteristic of an ratio scaled variable is that the origin is a fixed, unique zero-point.
Examples are: age, cost, number of consumers, etc.
Appropriate methods of analysis/statistics for ratio scales can be:
● Geometric average
● Range
● Coefficient of variation
+ all previous statistics of the nominal scale, ordinal scale and interval scale
ERRORS: RELIABILITY AND VALIDITY
Reliability and validity are both related to how well a method measures the concept of interest.
Reliability refers to the consistency of a measure (whether the results can be reproduced under the
same conditions). To check the reliability of the data the principle test-re-test can be used. For
example the same question is asked again later in the survey, then if the answers are the same the
data is reliable. Validity refers to the accuracy of a measure (whether the results really do represent
what they are supposed to measure).
STATISTICAL SIGNIFICANCE AND POWER
When conducting research, statistical tests are performed in order to test whether the formulated
hypotheses can be confirmed or not. While hypothesis testing the following may occur:
Statistical decision Reality
H0: no difference Ha: difference
H0: no difference 1–α β (type II error)
Ha: difference α (type I error) 1 – β (power)
A type I error (α) is the probability of the test showing statistical significance when it is not present
(‘false positive’). A type II error (β) is the probability of the test showing no statistical significance
when there is a difference. A specific cut-off can be used to reduce the chance on type I or type II
error. When dealing with type I error often a 5% cut-off level is used. Type II errors are not that
important for this course.
Power (1 – β) is the probability of the test showing statistical significance when it is present.
Power (1 – β) depends on:
● Alpha (+): If you are willing to accept are a higher type I error, the power of your
hypothesis testing will be higher.
● Effect size (+): If an effect - the difference being measured - is bigger in reality, you will
also have a higher chance of finding that difference with a given alpha and
sample size.
● Sample size (n) (+): The larger the sample size, the closer the sample size comes to the
population level. Which leads to a higher probability of finding a difference
that is present in reality. In other words, the power of finding a difference will
be higher.
Implications:
● Anticipate consequences of alpha, effect and n.
● Assess/incorporate power when interpreting results
2
, Demi van de Pol | Summary | Introduction to Research in Marketing | TISEM | Tilburg University | 2022
TYPES OF MULTIVARIATE METHODS
DEPENDENCE OR INTERDEPENDENCE TECHNIQUES
The main difference between dependence and interdependence techniques is the type of
relationship that is being examined. Therefore, you could ask yourself the question: What type of
relationship is being examined?
Dependence method Interdependence method
● ANOVA ● Factor analysis
● (Multiple) linear regression ● Cluster analysis
● Logistic regression
● Conjoint analysis
Dependence techniques are used to discover causal relationships. The goal is to find a link between
an outcome variable (result) and a set of variables that drive / cause that result (explanatory
variables).
When using dependence techniques one or more variables can be defined as dependent variables
influenced by one or more independent variables. The choice between the different dependence
techniques depends on the number of dependent variables involved in the analysis.
3
INTRODUCTION TO
RESEARCH IN
MARKETING
Demi van de Pol || Master Marketing Analytics & Marketing Management || Tilburg University || 2022
, Demi van de Pol | Summary | Introduction to Research in Marketing | TISEM | Tilburg University | 2022
CONTENT
This summary is written for the course “Introduction to Research in Marketing” during the semester
Spring-2022 and is part of both the master Marketing Analytics and the master Marketing
Management. The input for this summary consists of all the information given in the lectures.
LECTURE 1: Basics for multivariate analyses
DEFINITION MULTIVARIATE ANALYSIS
The term multivariate analysis refers to all statistical methods that simultaneously analyze multiple
measurements on each individual or object under investigation.
BASIC CONCEPTS FOR MULTIVARIATE ANALYSIS
MEASUREMENT SCALES
Measurement scales refer to how different variables can be collected / measured. There can be
made a distinction between non-metric scales (nominal and ordinal) and metric scales (interval and
ratio).
Non-metric measurement scale: Nominal
Nominal scaled variables are used to uniquely identify an object or person or to classify them. In
other words, the goal of nominal scaled variables is unique identification / definition or classification.
Appropriate methods of analysis / statistics for nominal scaled variables can be:
● Percentages (%)
● Mode
● Chi square tests
Non-metric measurement scale: Ordinal
Ordinal scaled variables are used to indicate ‘order’ / sequence. Examples of ordinal scaled variables
can be ranking, level of education, etc.
Appropriate methods of analysis/statistics for ordinal scales can be:
● Percentiles
● Median
● Rank correlation
+ all previous statistics of the nominal scale
Metric measurement scale: Interval
One important characteristic of an interval scaled variable is that the origin is arbitrary, meaning that
the origin not always starts at zero. Interval scaled variables allow for calculations.
Appropriate methods of analysis/statistics for interval scales can be:
● Arithmetic average
● Range
● Standard deviation
● Product-moment correlation
+ all previous statistics of the nominal scale and ordinal scale
1
, Demi van de Pol | Summary | Introduction to Research in Marketing | TISEM | Tilburg University | 2022
Metric measurement scale: Ratio
An important characteristic of an ratio scaled variable is that the origin is a fixed, unique zero-point.
Examples are: age, cost, number of consumers, etc.
Appropriate methods of analysis/statistics for ratio scales can be:
● Geometric average
● Range
● Coefficient of variation
+ all previous statistics of the nominal scale, ordinal scale and interval scale
ERRORS: RELIABILITY AND VALIDITY
Reliability and validity are both related to how well a method measures the concept of interest.
Reliability refers to the consistency of a measure (whether the results can be reproduced under the
same conditions). To check the reliability of the data the principle test-re-test can be used. For
example the same question is asked again later in the survey, then if the answers are the same the
data is reliable. Validity refers to the accuracy of a measure (whether the results really do represent
what they are supposed to measure).
STATISTICAL SIGNIFICANCE AND POWER
When conducting research, statistical tests are performed in order to test whether the formulated
hypotheses can be confirmed or not. While hypothesis testing the following may occur:
Statistical decision Reality
H0: no difference Ha: difference
H0: no difference 1–α β (type II error)
Ha: difference α (type I error) 1 – β (power)
A type I error (α) is the probability of the test showing statistical significance when it is not present
(‘false positive’). A type II error (β) is the probability of the test showing no statistical significance
when there is a difference. A specific cut-off can be used to reduce the chance on type I or type II
error. When dealing with type I error often a 5% cut-off level is used. Type II errors are not that
important for this course.
Power (1 – β) is the probability of the test showing statistical significance when it is present.
Power (1 – β) depends on:
● Alpha (+): If you are willing to accept are a higher type I error, the power of your
hypothesis testing will be higher.
● Effect size (+): If an effect - the difference being measured - is bigger in reality, you will
also have a higher chance of finding that difference with a given alpha and
sample size.
● Sample size (n) (+): The larger the sample size, the closer the sample size comes to the
population level. Which leads to a higher probability of finding a difference
that is present in reality. In other words, the power of finding a difference will
be higher.
Implications:
● Anticipate consequences of alpha, effect and n.
● Assess/incorporate power when interpreting results
2
, Demi van de Pol | Summary | Introduction to Research in Marketing | TISEM | Tilburg University | 2022
TYPES OF MULTIVARIATE METHODS
DEPENDENCE OR INTERDEPENDENCE TECHNIQUES
The main difference between dependence and interdependence techniques is the type of
relationship that is being examined. Therefore, you could ask yourself the question: What type of
relationship is being examined?
Dependence method Interdependence method
● ANOVA ● Factor analysis
● (Multiple) linear regression ● Cluster analysis
● Logistic regression
● Conjoint analysis
Dependence techniques are used to discover causal relationships. The goal is to find a link between
an outcome variable (result) and a set of variables that drive / cause that result (explanatory
variables).
When using dependence techniques one or more variables can be defined as dependent variables
influenced by one or more independent variables. The choice between the different dependence
techniques depends on the number of dependent variables involved in the analysis.
3
