Correlation
How strongly the pair of variables are related and change together.
Causation
Any change in the value of one variable will cause a change in the value of another.
An omitted variable causes…
both, the independent variable and the dependent variable. This is a classic “omitted variable bias”.
The two pillars of experimental research:
- Create systematic variation
in one variable and measure its effect on an outcome variable - Eliminate or mitigate unsystematic variation
through observable and unobservable attributes through randomization
Internal validity
How well you can ensure that the independent variable caused the dependent variable
Laboratory Experiment
- Great control, artificial environment created
- Participants are recruited
- Usually incentivized with money
- Real decisions with monetary consequences
Laboratory online experiment
Scenario Experiment
- Indirect environment. E.g. “imagine that… how would you..”
- Participants are recruited
- Imaginary decisions
- Monetary payment
Field experiment
- Natural environment. E.g. A/B testing, email campaign
- Participants are often unaware of being recruited
Quasi Experiment (also known as “natural experiment”)
Experiment without randomized controlled trial (RCT). E.g., natural experiment
A construct is…
an abstract concept that is specifically chosen (or “created”) to explain a given phenomenon
A variable is…
a measurable representation of a construct.
A proposition is…
is a tentative and conjectural relationship between constructs that is stated in a declarative form.
Hypotheses are…
are empirical formulations of propositions, stated as relationships between variables.
A theory is…
set of systematically interrelated constructs and propositions
intended to explain and predict a phenomenon or behavior of interest,
within certain boundary conditions and assumptions.
Experimental designs
- In its basic form, an experimental design is supposed to reduce the difference between treatment and control group to one factor, the manipulation.
- Experiments are (almost) the only way to keep other factors constant.
Between-participant/ posttest-only design
T Group: X-O
C Group: -O
This is the ‘classic design’
Within- and between-participant/ Pretest-posttest design
T Group: O-X-O
C Group: O- -O
Between-participant/ posttest-only design with control variables
T Group: C-X-O
C Group: C- -O
Covariance design
Factorial design
You have 4 groups and everyone gets a treatment
Two rare other group designs
- Solomon four group design
- Switched replication design
