Heterogeneity and Causality

2005/04/07 by Paul R. Rosenbaum, Paul R Rosenbaum · 213 citations
Mathematics · #Advanced Causal Inference Techniques #Artificial intelligence #Causal inference #Causality (physics) #Computer science #Econometrics #Inference #Key (lock) #Mathematics #Randomized experiment #Statistical Methods and Inference #Statistical Methods in Clinical Trials #Statistics

paper · doi:10.1198/000313005x42831

published in The American Statistician 59(2), 147-152 (Taylor & Francis)

openalex publication_date 2005/04/07 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/29

Abstract

Before R. A. Fisher introduced randomized experimentation, the literature on empirical methods emphasized reducing heterogeneity of experimental units as the key to inference about the effects caused by treatments. To what extent is heterogeneity relevant to causal inference when ethicalor practical constraints make random assignment infeasible?

Cited by

• Building Brand Assets: The Role of Trademark Rights
• Comparative Politics and the Synthetic Control Method
• How Much Is Minnesota Like Wisconsin? Assumptions and Counterfactuals in Causal Inference with Observational Data
• Re-Evaluating Strengthened-IV Designs: Asymptotic Efficiency, Bias Formula, and the Validity and Power of Sensitivity Analyses
• Discovering Effect Modification in an Observational Study of Surgical\n Mortality at Hospitals with Superior Nursing
• Are Randomized Controlled Trials the (G)old Standard? From Clinical Intelligence to Prescriptive Analytics. [europepmc]
• Empowering the poor: A field study of the social psychological consequences of receiving autonomy or dependency aid in Panama. [europepmc]
• Discovering cooperation: Endogenous change in international organizations. [europepmc]
• The R.O.A.D. to precision medicine. [europepmc]

Related

• A novel decomposition to explain heterogeneity in observational and randomized studies of causality
• Causal processes in psychology are heterogeneous.
• Causal Inference Using Potential Outcomes
• For objective causal inference, design trumps analysis
• Randomization does not imply unconfoundedness