Haphazard intentional allocation and rerandomization to improve covariate balance in experiments

Marcelo S. Lauretto, Rafael B. Stern, Kari L. Morgan, Margaret H. Clark, Julio M. Stern

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

In randomized experiments, a single random allocation can yield groups that differ meaningfully with respect to a given covariate. Furthermore, it is only feasible to use classical control procedures of allocation for a very modest number of covariates. As a response to this problem, Morgan and Rubin [11, 12] proposed an approach based on rerandomization (repeated randomization) to ensure that the final allocation obtained is balanced. However, despite the benefits of the rerandomization method, it has an exponential computational cost in the number of covariates, for fixed balance constraints. Here, we propose the use of haphazard intentional allocation, an alternative allocation method based on optimal balance of the covariates extended by random noise, see Lauretto et al. [7]. Our proposed method can be divided into a randomization and an optimization step. The randomization step consists of creating new (artificial) covariates according a specified distribution. The optimization step consists of finding the allocation that minimizes a linear combination of the imbalance in the original covariates and the imbalance in the artificial covariates. Numerical experiments on real and simulated data show a remarkable superiority of haphazard intentional allocation over the rerandomization method, both in terms of balance between groups and in terms of inference power.

Original languageEnglish (US)
Title of host publicationBayesian Inference and Maximum Entropy Methods in Science and Engineering
Subtitle of host publicationProceedings of the 36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016
EditorsGeert Verdoolaege
PublisherAmerican Institute of Physics Inc.
ISBN (Electronic)9780735415270
DOIs
StatePublished - Jun 9 2017
Event36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016 - Ghent, Belgium
Duration: Jul 10 2016Jul 15 2016

Publication series

NameAIP Conference Proceedings
Volume1853
ISSN (Print)0094-243X
ISSN (Electronic)1551-7616

Other

Other36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016
CountryBelgium
CityGhent
Period7/10/167/15/16

Fingerprint

optimization
random noise
inference
costs

All Science Journal Classification (ASJC) codes

  • Physics and Astronomy(all)

Cite this

Lauretto, M. S., Stern, R. B., Morgan, K. L., Clark, M. H., & Stern, J. M. (2017). Haphazard intentional allocation and rerandomization to improve covariate balance in experiments. In G. Verdoolaege (Ed.), Bayesian Inference and Maximum Entropy Methods in Science and Engineering: Proceedings of the 36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016 [050003] (AIP Conference Proceedings; Vol. 1853). American Institute of Physics Inc.. https://doi.org/10.1063/1.4985356
Lauretto, Marcelo S. ; Stern, Rafael B. ; Morgan, Kari L. ; Clark, Margaret H. ; Stern, Julio M. / Haphazard intentional allocation and rerandomization to improve covariate balance in experiments. Bayesian Inference and Maximum Entropy Methods in Science and Engineering: Proceedings of the 36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016. editor / Geert Verdoolaege. American Institute of Physics Inc., 2017. (AIP Conference Proceedings).
@inproceedings{02ff3b7eb03f414da6801ea7293e165d,
title = "Haphazard intentional allocation and rerandomization to improve covariate balance in experiments",
abstract = "In randomized experiments, a single random allocation can yield groups that differ meaningfully with respect to a given covariate. Furthermore, it is only feasible to use classical control procedures of allocation for a very modest number of covariates. As a response to this problem, Morgan and Rubin [11, 12] proposed an approach based on rerandomization (repeated randomization) to ensure that the final allocation obtained is balanced. However, despite the benefits of the rerandomization method, it has an exponential computational cost in the number of covariates, for fixed balance constraints. Here, we propose the use of haphazard intentional allocation, an alternative allocation method based on optimal balance of the covariates extended by random noise, see Lauretto et al. [7]. Our proposed method can be divided into a randomization and an optimization step. The randomization step consists of creating new (artificial) covariates according a specified distribution. The optimization step consists of finding the allocation that minimizes a linear combination of the imbalance in the original covariates and the imbalance in the artificial covariates. Numerical experiments on real and simulated data show a remarkable superiority of haphazard intentional allocation over the rerandomization method, both in terms of balance between groups and in terms of inference power.",
author = "Lauretto, {Marcelo S.} and Stern, {Rafael B.} and Morgan, {Kari L.} and Clark, {Margaret H.} and Stern, {Julio M.}",
year = "2017",
month = "6",
day = "9",
doi = "10.1063/1.4985356",
language = "English (US)",
series = "AIP Conference Proceedings",
publisher = "American Institute of Physics Inc.",
editor = "Geert Verdoolaege",
booktitle = "Bayesian Inference and Maximum Entropy Methods in Science and Engineering",

}

Lauretto, MS, Stern, RB, Morgan, KL, Clark, MH & Stern, JM 2017, Haphazard intentional allocation and rerandomization to improve covariate balance in experiments. in G Verdoolaege (ed.), Bayesian Inference and Maximum Entropy Methods in Science and Engineering: Proceedings of the 36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016., 050003, AIP Conference Proceedings, vol. 1853, American Institute of Physics Inc., 36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016, Ghent, Belgium, 7/10/16. https://doi.org/10.1063/1.4985356

Haphazard intentional allocation and rerandomization to improve covariate balance in experiments. / Lauretto, Marcelo S.; Stern, Rafael B.; Morgan, Kari L.; Clark, Margaret H.; Stern, Julio M.

Bayesian Inference and Maximum Entropy Methods in Science and Engineering: Proceedings of the 36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016. ed. / Geert Verdoolaege. American Institute of Physics Inc., 2017. 050003 (AIP Conference Proceedings; Vol. 1853).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Haphazard intentional allocation and rerandomization to improve covariate balance in experiments

AU - Lauretto, Marcelo S.

AU - Stern, Rafael B.

AU - Morgan, Kari L.

AU - Clark, Margaret H.

AU - Stern, Julio M.

PY - 2017/6/9

Y1 - 2017/6/9

N2 - In randomized experiments, a single random allocation can yield groups that differ meaningfully with respect to a given covariate. Furthermore, it is only feasible to use classical control procedures of allocation for a very modest number of covariates. As a response to this problem, Morgan and Rubin [11, 12] proposed an approach based on rerandomization (repeated randomization) to ensure that the final allocation obtained is balanced. However, despite the benefits of the rerandomization method, it has an exponential computational cost in the number of covariates, for fixed balance constraints. Here, we propose the use of haphazard intentional allocation, an alternative allocation method based on optimal balance of the covariates extended by random noise, see Lauretto et al. [7]. Our proposed method can be divided into a randomization and an optimization step. The randomization step consists of creating new (artificial) covariates according a specified distribution. The optimization step consists of finding the allocation that minimizes a linear combination of the imbalance in the original covariates and the imbalance in the artificial covariates. Numerical experiments on real and simulated data show a remarkable superiority of haphazard intentional allocation over the rerandomization method, both in terms of balance between groups and in terms of inference power.

AB - In randomized experiments, a single random allocation can yield groups that differ meaningfully with respect to a given covariate. Furthermore, it is only feasible to use classical control procedures of allocation for a very modest number of covariates. As a response to this problem, Morgan and Rubin [11, 12] proposed an approach based on rerandomization (repeated randomization) to ensure that the final allocation obtained is balanced. However, despite the benefits of the rerandomization method, it has an exponential computational cost in the number of covariates, for fixed balance constraints. Here, we propose the use of haphazard intentional allocation, an alternative allocation method based on optimal balance of the covariates extended by random noise, see Lauretto et al. [7]. Our proposed method can be divided into a randomization and an optimization step. The randomization step consists of creating new (artificial) covariates according a specified distribution. The optimization step consists of finding the allocation that minimizes a linear combination of the imbalance in the original covariates and the imbalance in the artificial covariates. Numerical experiments on real and simulated data show a remarkable superiority of haphazard intentional allocation over the rerandomization method, both in terms of balance between groups and in terms of inference power.

UR - http://www.scopus.com/inward/record.url?scp=85021414531&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85021414531&partnerID=8YFLogxK

U2 - 10.1063/1.4985356

DO - 10.1063/1.4985356

M3 - Conference contribution

AN - SCOPUS:85021414531

T3 - AIP Conference Proceedings

BT - Bayesian Inference and Maximum Entropy Methods in Science and Engineering

A2 - Verdoolaege, Geert

PB - American Institute of Physics Inc.

ER -

Lauretto MS, Stern RB, Morgan KL, Clark MH, Stern JM. Haphazard intentional allocation and rerandomization to improve covariate balance in experiments. In Verdoolaege G, editor, Bayesian Inference and Maximum Entropy Methods in Science and Engineering: Proceedings of the 36th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2016. American Institute of Physics Inc. 2017. 050003. (AIP Conference Proceedings). https://doi.org/10.1063/1.4985356