Randomization-based causal inference from split-plot designs

Anqi Zhao, Peng Ding, Rahul Mukerjee, Tirthankar Dasgupta

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

Under the potential outcomes framework, we propose a randomization based estimation procedure for causal inference from split-plot designs, with special emphasis on 22 designs that naturally arise in many social, behavioral and biomedical experiments. Point estimators of factorial effects are obtained and their sampling variances are derived in closed form as linear combinations of the between- and within-group covariances of the potential outcomes. Results are compared to those under complete randomization as measures of design efficiency. Conservative estimators of these sampling variances are proposed. Connection of the randomization-based approach to inference based on the linear mixed effects model is explored. Results on sampling variances of point estimators and their estimators are extended to general split-plot designs. The superiority over existing model-based alternatives in frequency coverage properties is reported under a variety of simulation settings for both binary and continuous outcomes.

Original languageEnglish (US)
Pages (from-to)1876-1903
Number of pages28
JournalAnnals of Statistics
Volume46
Issue number5
DOIs
StatePublished - Oct 1 2018

Fingerprint

Split-plot Design
Causal Inference
Randomisation
Potential Outcomes
Estimator
Design Efficiency
Linear Mixed Effects Model
Factorial
Linear Combination
Closed-form
Coverage
Model-based
Binary
Randomization
Causal inference
Alternatives
Experiment
Sampling
Simulation
Potential outcomes

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Keywords

  • Between-whole-plot additivity
  • Model-based inference
  • Neymanian inference
  • Potential outcomes framework
  • Projection matrix
  • Within-whole-plot additivity

Cite this

Zhao, Anqi ; Ding, Peng ; Mukerjee, Rahul ; Dasgupta, Tirthankar. / Randomization-based causal inference from split-plot designs. In: Annals of Statistics. 2018 ; Vol. 46, No. 5. pp. 1876-1903.
@article{3b26ffa051354f39ad6f2e29da3582b6,
title = "Randomization-based causal inference from split-plot designs",
abstract = "Under the potential outcomes framework, we propose a randomization based estimation procedure for causal inference from split-plot designs, with special emphasis on 22 designs that naturally arise in many social, behavioral and biomedical experiments. Point estimators of factorial effects are obtained and their sampling variances are derived in closed form as linear combinations of the between- and within-group covariances of the potential outcomes. Results are compared to those under complete randomization as measures of design efficiency. Conservative estimators of these sampling variances are proposed. Connection of the randomization-based approach to inference based on the linear mixed effects model is explored. Results on sampling variances of point estimators and their estimators are extended to general split-plot designs. The superiority over existing model-based alternatives in frequency coverage properties is reported under a variety of simulation settings for both binary and continuous outcomes.",
keywords = "Between-whole-plot additivity, Model-based inference, Neymanian inference, Potential outcomes framework, Projection matrix, Within-whole-plot additivity",
author = "Anqi Zhao and Peng Ding and Rahul Mukerjee and Tirthankar Dasgupta",
year = "2018",
month = "10",
day = "1",
doi = "https://doi.org/10.1214/17-AOS1605",
language = "English (US)",
volume = "46",
pages = "1876--1903",
journal = "Annals of Statistics",
issn = "0090-5364",
publisher = "Institute of Mathematical Statistics",
number = "5",

}

Randomization-based causal inference from split-plot designs. / Zhao, Anqi; Ding, Peng; Mukerjee, Rahul; Dasgupta, Tirthankar.

In: Annals of Statistics, Vol. 46, No. 5, 01.10.2018, p. 1876-1903.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Randomization-based causal inference from split-plot designs

AU - Zhao, Anqi

AU - Ding, Peng

AU - Mukerjee, Rahul

AU - Dasgupta, Tirthankar

PY - 2018/10/1

Y1 - 2018/10/1

N2 - Under the potential outcomes framework, we propose a randomization based estimation procedure for causal inference from split-plot designs, with special emphasis on 22 designs that naturally arise in many social, behavioral and biomedical experiments. Point estimators of factorial effects are obtained and their sampling variances are derived in closed form as linear combinations of the between- and within-group covariances of the potential outcomes. Results are compared to those under complete randomization as measures of design efficiency. Conservative estimators of these sampling variances are proposed. Connection of the randomization-based approach to inference based on the linear mixed effects model is explored. Results on sampling variances of point estimators and their estimators are extended to general split-plot designs. The superiority over existing model-based alternatives in frequency coverage properties is reported under a variety of simulation settings for both binary and continuous outcomes.

AB - Under the potential outcomes framework, we propose a randomization based estimation procedure for causal inference from split-plot designs, with special emphasis on 22 designs that naturally arise in many social, behavioral and biomedical experiments. Point estimators of factorial effects are obtained and their sampling variances are derived in closed form as linear combinations of the between- and within-group covariances of the potential outcomes. Results are compared to those under complete randomization as measures of design efficiency. Conservative estimators of these sampling variances are proposed. Connection of the randomization-based approach to inference based on the linear mixed effects model is explored. Results on sampling variances of point estimators and their estimators are extended to general split-plot designs. The superiority over existing model-based alternatives in frequency coverage properties is reported under a variety of simulation settings for both binary and continuous outcomes.

KW - Between-whole-plot additivity

KW - Model-based inference

KW - Neymanian inference

KW - Potential outcomes framework

KW - Projection matrix

KW - Within-whole-plot additivity

UR - http://www.scopus.com/inward/record.url?scp=85052652941&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85052652941&partnerID=8YFLogxK

U2 - https://doi.org/10.1214/17-AOS1605

DO - https://doi.org/10.1214/17-AOS1605

M3 - Article

VL - 46

SP - 1876

EP - 1903

JO - Annals of Statistics

JF - Annals of Statistics

SN - 0090-5364

IS - 5

ER -