The use of Thompson sampling to increase estimation precision

Research output: Contribution to journalArticleScientificpeer-review

Abstract

In this article, we consider a sequential sampling scheme for efficient estimation of the difference between the means of two independent treatments when the population variances are unequal across groups. The sampling scheme proposed is based on a solution to bandit problems called Thompson sampling. While this approach is most often used to maximize the cumulative payoff over competing treatments, we show that the same method can also be used to balance exploration and exploitation when the aim of the experimenter is to efficiently increase estimation precision. We introduce this novel design optimization method and, by simulation, show its effectiveness.
Keywords: Design optimization, Thompson sampling, Bandit problems
Original languageEnglish
Pages (from-to)409-423
JournalBehavior Research Methods
Volume47
Issue number2
DOIs
Publication statusPublished - 2015

Keywords

  • Design optimization
  • Thompson sampling
  • Bandit problems

Cite this

@article{04e93c8723eb440c9db30cfd5ba3ef99,
title = "The use of Thompson sampling to increase estimation precision",
abstract = "In this article, we consider a sequential sampling scheme for efficient estimation of the difference between the means of two independent treatments when the population variances are unequal across groups. The sampling scheme proposed is based on a solution to bandit problems called Thompson sampling. While this approach is most often used to maximize the cumulative payoff over competing treatments, we show that the same method can also be used to balance exploration and exploitation when the aim of the experimenter is to efficiently increase estimation precision. We introduce this novel design optimization method and, by simulation, show its effectiveness.Keywords: Design optimization, Thompson sampling, Bandit problems",
keywords = "Design optimization, Thompson sampling, Bandit problems",
author = "M.C. Kaptein",
year = "2015",
doi = "10.3758/s13428-014-0480-0",
language = "English",
volume = "47",
pages = "409--423",
journal = "Behavior Research Methods",
issn = "1554-351X",
publisher = "Springer",
number = "2",

}

The use of Thompson sampling to increase estimation precision. / Kaptein, M.C.

In: Behavior Research Methods, Vol. 47, No. 2, 2015, p. 409-423.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - The use of Thompson sampling to increase estimation precision

AU - Kaptein, M.C.

PY - 2015

Y1 - 2015

N2 - In this article, we consider a sequential sampling scheme for efficient estimation of the difference between the means of two independent treatments when the population variances are unequal across groups. The sampling scheme proposed is based on a solution to bandit problems called Thompson sampling. While this approach is most often used to maximize the cumulative payoff over competing treatments, we show that the same method can also be used to balance exploration and exploitation when the aim of the experimenter is to efficiently increase estimation precision. We introduce this novel design optimization method and, by simulation, show its effectiveness.Keywords: Design optimization, Thompson sampling, Bandit problems

AB - In this article, we consider a sequential sampling scheme for efficient estimation of the difference between the means of two independent treatments when the population variances are unequal across groups. The sampling scheme proposed is based on a solution to bandit problems called Thompson sampling. While this approach is most often used to maximize the cumulative payoff over competing treatments, we show that the same method can also be used to balance exploration and exploitation when the aim of the experimenter is to efficiently increase estimation precision. We introduce this novel design optimization method and, by simulation, show its effectiveness.Keywords: Design optimization, Thompson sampling, Bandit problems

KW - Design optimization

KW - Thompson sampling

KW - Bandit problems

U2 - 10.3758/s13428-014-0480-0

DO - 10.3758/s13428-014-0480-0

M3 - Article

VL - 47

SP - 409

EP - 423

JO - Behavior Research Methods

JF - Behavior Research Methods

SN - 1554-351X

IS - 2

ER -