Experimental Research Design
Experimental research design connects probability, statistical inference, and causal identification. The central question is not merely whether two variables are associated, but whether a treatment caused a change in an outcome.
Use this course in this order:
- Foundations for Experimental Research
- Regression Models for Experimental Research
- Randomized Experiments
- Observational Causal Inference
- Experimental Research Design Formula Sheet
Practice after each section:
- Foundations Exam Workshop
- Regression Exam Workshop
- Randomized Experiments Exam Workshop
- Observational Causal Inference Exam Workshop
Course map
probability and sampling distributions
-> tests, confidence intervals, and ANOVA
-> regression as a model and adjustment tool
-> potential outcomes and the missing counterfactual
-> randomized assignment
-> Neyman and Fisher inference
-> blocking, pairing, and regression adjustment
-> observational identification under assumptions
-> propensity scores, weighting, and matching
The three layers to keep separate
- Design determines how treatment is assigned and which comparisons are credible.
- Estimation defines how the treatment effect is calculated, such as a difference in means.
- Inference quantifies uncertainty or tests a hypothesis under the assignment or sampling process.
Random assignment supplies identification by design. In observational studies, identification instead depends on assumptions such as Unconfoundedness and Overlap.
Core notation
- $i=1,\ldots,N$ indexes units.
- $W_i\in{0,1}$ is the treatment indicator.
- $Y_i(1)$ and $Y_i(0)$ are potential outcomes.
- $Y_i^{obs}$ is the observed outcome.
- $N_1$ and $N_0$ are the numbers assigned to treatment and control.
- $\tau_i=Y_i(1)-Y_i(0)$ is an individual treatment effect.
- $\hat\tau$ is an estimated average treatment effect.
Exam-first route
- Identify the design: independent, paired, stratified, randomized, or observational.
- State the estimand and assumptions.
- Select the estimator and its standard error.
- State the null hypothesis and reference distribution.
- Compute an interval or $p$-value and interpret it in context.
- Use Experimental Research Design Formula Sheet for rapid recall.
Connections
- Inferential Statistics develops the sampling-based tools used throughout the course.
- Econometrics develops regression and causal interpretation in greater breadth.
- Behavioral Economics supplies applications in which experimental design is central.
- Linear Regression and Logistic Regression give prediction-focused treatments of two models reused here for inference.