Regression Models for Experimental Research
Regression summarizes outcome differences, adjusts for covariates, and represents heterogeneous treatment effects. Its causal meaning comes from the research design and identifying assumptions.
See
-
Linear Regression for Experimental Research
- Simple and multiple OLS, residual variance, coefficient tests, and $R^2$.
-
Indicator Variables and Interactions
- Group differences and group-specific slopes.
-
Binary Outcome Models for Experimental Research
- Linear probability and logistic models.
-
Regression Adjustment in Experiments
- Treatment-effect estimation with and without covariates and interactions.
Interpretation rule
A regression coefficient is a model-based conditional association by default. It becomes a causal effect only when assignment or assumptions make the relevant comparison causal. In a randomized experiment, treatment assignment provides that foundation. In an observational study, see Unconfoundedness and Overlap.
Worked problems and practice
Regression Exam Workshop develops treatment interactions, coefficient inference, $R^2$, logistic probabilities, and odds ratios through multi-part exam problems.
Connections
- ANOVA for Experimental Research is regression with group indicators.
- Randomized Assignment explains when the coefficient on treatment recovers a causal effect.
- Propensity Scores often uses logistic regression to model assignment rather than outcomes.