Introduction to research design:
Quantitative research = research approach that examines concepts in terms of
amount, intensity, or frequency
Qualitative research = research approach that examines concepts in terms of
their meaning and interpretation in specific contexts on inquiry
Two ways of performing qualitative research:
1. Variance theory
= explaining strategic change with a variance model
(explains outcomes by focusing on how different levels of independent variables
relate to variations in a dependent variable)
2. Process theory
= explaining strategic change with a process model
(explains how an entity changes and develops by focusing on the sequence of
events and activities over time)
Forms of reasoning:
1. Abduction: inference to a cause/case
-> used to frame problems and generate solutions/theories
(Example: that cyclist must have a flat tire)
2. Deduction: inference to a result
-> used for testing solutions/theories
(Example: the cyclist is fixing the problem / repairing his bike)
3. Induction: inference to a rule
-> involves generalizing from a sample to a population
(Example: on average, cyclists doing similar actions by the side of the road are
likely repairing their bicycles)
We predict, confirm, and disconfirm through deduction, generalize through
induction, and theorize through abduction
Abduce a conjecture -> deduce a consequence -> induce a rule
Lecture 1:
The process of building theory:
1. Observe, describe & measure the phenomena (Constructs)
2. Categorization based upon attributes of phenomena (Frameworks &
Typologies)
3. Statements of association (Models)
Deductive process: 3 -> 1 (using a model to predict the phenomena)
Inductive process: 1 -> 3 (using the phenomena to create a model)
Lecture 2:
Internal validity -> given that there is a relationship, is the relationship a causal
one?
* Speaks to the validity of the research itself
,* Selection bias, and alternative explanations can threaten internal validity
Construct validity -> how well did you translate your construct into a
functioning and operating reality
External validity -> if causal relationships exist, how generalizable is this
relationship
The key assumptions of linear regression:
1. Linearity (a linear relationship between X and Y)
-> if nonlinear, you need to transform the variable to make it linear
2. Homoscedasticity (an equal variance)
3. Normality of Errors (normality of error distribution)
4. Independence of Errors (independence of observations, includes ‘no
autocorrelation’)
5. No Multicollinearity (predictors are not correlated with each other)
-> calculate the VIF (variance influation factor), if the value is less than 10, there
is no multicollinearity
6. No endogeneity (no correlations between predictors and errors)
Linear Regression
= specifies the conditional mean of a response variable ‘y’ as a linear function of
‘k’ independent variables
Simple linear regression -> one independent variable
Multiple linear regression -> multiple independent variables
Standard error = estimates the sampling distribution of the coefficient in the
population
Lecture 3:
Potential outcome model
= defines causal effects by comparing what would have happened to an
individual under treatment versus no treatment, even though only one outcome
can be observed for any given individual
Experiments:
Types of experiments:
1. Laboratory experiments
▪ usually controlled and incentivized
▪ imposed set of rules -> internal validity
▪ standard subject pool
▪ abstract framing
2. Field experiments
▪ artefactual (uses a non-standard subject pool)
▪ framed: provides a specific framing, or context
▪ natural: environment is one in which subjects naturally undertake tasks and
where subjects do now know they are in an experiments
▪ focus: external validity
, 3. Stated Preference experiments
▪ hypothetical choices
▪ conjoint analysis
Why do we use experiments?
1. Self-serving bias of people / people are not always aware of our own mistakes
2. Secondary data is inferior to experiments
Focus variables = the effect of the variables in which you are interested
Nuisance variables = are of no direct interest but may effect results
Confounding = the effect of two or more variables
Full factorial design = designing an experiment where all combinations of
variables are tested, a treatment is conducted for each combination
-> this provides the cleanest evidence for the effect of each variable, but can get
quite expensive if there are many variables
Between-subject design = independent group of participants in each
treatment
Within-subject design = the same subject experiences more than one
treatment
+ multiple observations per participant
+ stronger statistical or paired tests
- order effects
-> solution for this is to randomize the order of treatments
Stated Preference experiments
= a data collection method to examine the trade-offs people make when making
a choice
▪ hypothetical alternatives are constructed and presented to respondents
-> respondents are either asked to give a rating to various options or to pick their
preferred option
The estimation of the utility function depends on what type of response type is
used in the experiment
1. Ratings: Regression
▪ rating should be an internal level measurement
▪ DV: ratings of hypotherital alternatives
▪ IV: coded attribute levels
2. Choices: Multinomial-logit model (MNL)
▪ DV: 1 (alternative was chosen) or 0 (alternative was not chosen)
▪ IV: coded attribute levels
-> choice-based models are the preferred academic standard
Three different measurement tasks are generally distinguished:
1. Choice tasks
2. Rating tasks
3. Ranking tasks
Secondary data and archival research:
Quantitative research = research approach that examines concepts in terms of
amount, intensity, or frequency
Qualitative research = research approach that examines concepts in terms of
their meaning and interpretation in specific contexts on inquiry
Two ways of performing qualitative research:
1. Variance theory
= explaining strategic change with a variance model
(explains outcomes by focusing on how different levels of independent variables
relate to variations in a dependent variable)
2. Process theory
= explaining strategic change with a process model
(explains how an entity changes and develops by focusing on the sequence of
events and activities over time)
Forms of reasoning:
1. Abduction: inference to a cause/case
-> used to frame problems and generate solutions/theories
(Example: that cyclist must have a flat tire)
2. Deduction: inference to a result
-> used for testing solutions/theories
(Example: the cyclist is fixing the problem / repairing his bike)
3. Induction: inference to a rule
-> involves generalizing from a sample to a population
(Example: on average, cyclists doing similar actions by the side of the road are
likely repairing their bicycles)
We predict, confirm, and disconfirm through deduction, generalize through
induction, and theorize through abduction
Abduce a conjecture -> deduce a consequence -> induce a rule
Lecture 1:
The process of building theory:
1. Observe, describe & measure the phenomena (Constructs)
2. Categorization based upon attributes of phenomena (Frameworks &
Typologies)
3. Statements of association (Models)
Deductive process: 3 -> 1 (using a model to predict the phenomena)
Inductive process: 1 -> 3 (using the phenomena to create a model)
Lecture 2:
Internal validity -> given that there is a relationship, is the relationship a causal
one?
* Speaks to the validity of the research itself
,* Selection bias, and alternative explanations can threaten internal validity
Construct validity -> how well did you translate your construct into a
functioning and operating reality
External validity -> if causal relationships exist, how generalizable is this
relationship
The key assumptions of linear regression:
1. Linearity (a linear relationship between X and Y)
-> if nonlinear, you need to transform the variable to make it linear
2. Homoscedasticity (an equal variance)
3. Normality of Errors (normality of error distribution)
4. Independence of Errors (independence of observations, includes ‘no
autocorrelation’)
5. No Multicollinearity (predictors are not correlated with each other)
-> calculate the VIF (variance influation factor), if the value is less than 10, there
is no multicollinearity
6. No endogeneity (no correlations between predictors and errors)
Linear Regression
= specifies the conditional mean of a response variable ‘y’ as a linear function of
‘k’ independent variables
Simple linear regression -> one independent variable
Multiple linear regression -> multiple independent variables
Standard error = estimates the sampling distribution of the coefficient in the
population
Lecture 3:
Potential outcome model
= defines causal effects by comparing what would have happened to an
individual under treatment versus no treatment, even though only one outcome
can be observed for any given individual
Experiments:
Types of experiments:
1. Laboratory experiments
▪ usually controlled and incentivized
▪ imposed set of rules -> internal validity
▪ standard subject pool
▪ abstract framing
2. Field experiments
▪ artefactual (uses a non-standard subject pool)
▪ framed: provides a specific framing, or context
▪ natural: environment is one in which subjects naturally undertake tasks and
where subjects do now know they are in an experiments
▪ focus: external validity
, 3. Stated Preference experiments
▪ hypothetical choices
▪ conjoint analysis
Why do we use experiments?
1. Self-serving bias of people / people are not always aware of our own mistakes
2. Secondary data is inferior to experiments
Focus variables = the effect of the variables in which you are interested
Nuisance variables = are of no direct interest but may effect results
Confounding = the effect of two or more variables
Full factorial design = designing an experiment where all combinations of
variables are tested, a treatment is conducted for each combination
-> this provides the cleanest evidence for the effect of each variable, but can get
quite expensive if there are many variables
Between-subject design = independent group of participants in each
treatment
Within-subject design = the same subject experiences more than one
treatment
+ multiple observations per participant
+ stronger statistical or paired tests
- order effects
-> solution for this is to randomize the order of treatments
Stated Preference experiments
= a data collection method to examine the trade-offs people make when making
a choice
▪ hypothetical alternatives are constructed and presented to respondents
-> respondents are either asked to give a rating to various options or to pick their
preferred option
The estimation of the utility function depends on what type of response type is
used in the experiment
1. Ratings: Regression
▪ rating should be an internal level measurement
▪ DV: ratings of hypotherital alternatives
▪ IV: coded attribute levels
2. Choices: Multinomial-logit model (MNL)
▪ DV: 1 (alternative was chosen) or 0 (alternative was not chosen)
▪ IV: coded attribute levels
-> choice-based models are the preferred academic standard
Three different measurement tasks are generally distinguished:
1. Choice tasks
2. Rating tasks
3. Ranking tasks
Secondary data and archival research:
