Causal inference and experimentation

Randomised controlled trials is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

A/B testing is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Multivariate testing is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Split-cell experiments is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Factorial experiments is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Fractional-factorial experiments is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Quasi-experimental design is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Difference-in-differences estimates an intervention effect by comparing changes over time between treated and comparison groups. Its credibility relies heavily on the parallel-trends assumption and the absence of differential shocks.

Regression discontinuity is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Instrumental variables is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Propensity-score matching pairs treated and untreated observations with similar estimated treatment probabilities. It can balance observed covariates but cannot remove bias from unobserved confounding.

Propensity-score weighting is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Coarsened exact matching is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Synthetic control constructs a weighted combination of comparison units that approximates the treated unit before an intervention. Post-intervention divergence is interpreted against placebo and robustness checks.

Interrupted time series is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Uplift modelling is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Treatment-effect modelling is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Heterogeneous treatment effects is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Causal forests is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Double machine learning is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Mediation analysis is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Incrementality testing is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Geo experiments is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Matched-market testing is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Marketing-mix experiments is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Holdout-market tests is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Causal impact analysis is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Bayesian experiments is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.

Sequential testing is a method within causal inference and experimentation. Experimental and quasi-experimental methods designed to estimate what changed because of an intervention rather than alongside it. Its usefulness depends on data structure, assumptions, validation and whether the output answers the intended decision.

Multi-armed bandits is a statistical or analytical concept within causal inference and experimentation. It should be selected for the data-generating process and decision question rather than because software makes it available.