A paradox from randomization-based causal inference
From MaRDI portal
Publication:1750243
Abstract: Under the potential outcomes framework, causal effects are defined as comparisons between potential outcomes under treatment and control. To infer causal effects from randomized experiments, Neyman proposed to test the null hypothesis of zero average causal effect (Neyman's null), and Fisher proposed to test the null hypothesis of zero individual causal effect (Fisher's null). Although the subtle difference between Neyman's null and Fisher's null has caused lots of controversies and confusions for both theoretical and practical statisticians, a careful comparison between the two approaches has been lacking in the literature for more than eighty years. We fill in this historical gap by making a theoretical comparison between them and highlighting an intriguing paradox that has not been recognized by previous researchers. Logically, Fisher's null implies Neyman's null. It is therefore surprising that, in actual completely randomized experiments, rejection of Neyman's null does not imply rejection of Fisher's null for many realistic situations, including the case with constant causal effect. Furthermore, we show that this paradox also exists in other commonly-used experiments, such as stratified experiments, matched-pair experiments, and factorial experiments. Asymptotic analyses, numerical examples, and real data examples all support this surprising phenomenon. Besides its historical and theoretical importance, this paradox also leads to useful practical implications for modern researchers.
Recommendations
Cites work
- A closer look at testing the ``no-treatment-effect hypothesis in a comparative experiment
- A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity
- A paradox from randomization-based causal inference
- Agnostic notes on regression adjustments to experimental data: reexamining Freedman's critique
- An introduction to copulas.
- Asymptotic permutation tests in general factorial designs
- Causal inference for statistics, social, and biomedical sciences. An introduction
- Causal inference from 2K factorial designs by using potential outcomes
- Comment on J. Neyman and causal inference in experiments and observational studies: On the application of probability theory to agricultural experiments. Essay on principles. Section 9 [Ann. Agric. Sci. 10 (1923), 1--51]
- Comments on the Neyman-Fisher controversy and its consequences
- Conditional rank tests for the two-sample problem under random censorship
- Design and Analysis of Experiments, Second Edition
- Exact and asymptotically robust permutation tests
- Experiments. Planning, analysis and optimization.
- scientific article; zbMATH DE number 3141625 (Why is no real title available?)
- scientific article; zbMATH DE number 3166683 (Why is no real title available?)
- scientific article; zbMATH DE number 3817660 (Why is no real title available?)
- scientific article; zbMATH DE number 1249686 (Why is no real title available?)
- scientific article; zbMATH DE number 627766 (Why is no real title available?)
- scientific article; zbMATH DE number 3213198 (Why is no real title available?)
- scientific article; zbMATH DE number 3318338 (Why is no real title available?)
- scientific article; zbMATH DE number 3336465 (Why is no real title available?)
- scientific article; zbMATH DE number 3029212 (Why is no real title available?)
- scientific article; zbMATH DE number 3030169 (Why is no real title available?)
- scientific article; zbMATH DE number 3076948 (Why is no real title available?)
- Is regression adjustment supported by the Neyman model for causal inference?
- Mostly harmless econometrics. An empiricist's companion.
- Non-Additives in a Latin Square Design
- Observational studies.
- On equivalencies between design-based and regression-based variance estimators for randomized experiments
- On regression adjustments to experimental data
- On the application of probability theory to agricultural experiments. Essay on principles. Section 9. Translated from the Polish and edited by D. M. Dąbrowska and T. P. Speed
- ON THE z-TEST IN RANDOMIZED BLOCKS AND LATIN SQUARES
- Outline of a Theory of Statistical Estimation Based on the Classical Theory of Probability
- Randomization inference for treatment effect variation
- Randomization tests. With CD-ROM.
- Reconsidering the Fundamental Contributions of Fisher and Neyman on Experimentation and Sampling
- Sharp bounds on the variance in randomized experiments
- Studentized permutation tests for non-i.i.d. hypotheses and the generalized Behrens-Fisher problem
- THE INTERPRETATION OF THE EFFECTS OF NON-ADDITIVITY IN THE LATIN SQUARE
- The Large-Sample Power of Tests Based on Permutations of Observations
- THE MEANING OF A SIGNIFICANCE LEVEL
Cited in
(39)- Humphreys's paradox and the interpretation of inverse conditional propensities
- Generalizing distance covariance to measure and test multivariate mutual dependence via complete and incomplete V-statistics
- Rank tests in unmatched clustered randomized trials applied to a study of teacher training
- A paradox from randomization-based causal inference
- Understanding Ding's apparent paradox
- Randomization-based tests for ``no treatment effects
- Inference from randomized (factorial) experiments
- An apparent paradox explained
- Rejoinder: A paradox from randomization-based causal inference
- On the power of conditional independence testing under model-X
- Detecting heterogeneous treatment effects with instrumental variables and application to the Oregon Health Insurance Experiment
- A closer look at testing the ``no-treatment-effect hypothesis in a comparative experiment
- Randomization-based test for censored outcomes: a new look at the logrank test
- On randomization-based and regression-based inferences for 2K factorial designs
- Causal inference through a witness protection program
- Using the potential outcome framework to estimate optimal sample size for cluster randomized trials: a simulation-based algorithm
- Model‐free causal inference of binary experimental data
- On finite-population Bayesian inferences for 2K factorial designs with binary outcomes
- Studentized sensitivity analysis for the sample average treatment effect in paired observational studies
- Resurrecting the Third Variable: A Critique of Pearl's Causal Analysis of Simpson's Paradox
- Correction
- Minimax Efficient Random Experimental Design Strategies With Application to Model-Robust Design for Prediction
- Randomization Tests for Weak Null Hypotheses in Randomized Experiments
- Improved Neymanian analysis for 2K factorial designs with binary outcomes
- Testing Weak Nulls in Matched Observational Studies
- Inference in Experiments With Matched Pairs
- Testing for the presence of treatment effect under selection on observables
- Distributed testing on mutual independence of massive multivariate data
- Population interference in panel experiments
- Is Fisher inference inferior to Neyman inference for policy analysis?
- The 41\textsuperscript{st} Fisher Memorial Lecture. From Fisher to CARA: the evolution of randomization and randomization-based inference
- Testing for causal effect for binary data when propensity scores are estimated through Bayesian networks
- Sharp variance estimator and causal bootstrap in stratified randomized experiments
- Sequential Monte Carlo testing by betting
- Sharp nonparametric bounds and randomization inference for treatment effects on an ordinal outcome
- Causal estimands and confidence intervals associated with Wilcoxon-Mann-Whitney tests in randomized experiments
- Decomposing Treatment Effect Variation
- A causal bootstrap
- The Frisch-Waugh-Lovell theorem for standard errors
This page was built for publication: A paradox from randomization-based causal inference
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1750243)
