Optimal design of experiments on connected units with application to social networks

Ben M. Parker; Steven G. Gilmour; John Alexander Schormans

doi:10.1111/rssc.12170

Optimal design of experiments on connected units with application to social networks

Ben M. Parker, Steven G. Gilmour, John Alexander Schormans

Research output: Contribution to journal › Article › peer-review

13 Citations (Scopus)

206 Downloads (Pure)

Abstract

When experiments are performed on social networks, it is difficult to justify the usual assumption of treatment-unit additivity, due to the connections between actors in the network. We investigate how connections between experimental units affect the design of experiments on those experimental units. Specifically, where we have unstructured treatments, whose effects propagate according to a linear network effects model which we introduce, we show that optimal designs are no longer necessarily balanced; we further demonstrate how experiments which do not take a network effect into account can lead to much higher variance than necessary and/or a large bias. We show the use of this methodology in a very wide range of experiments in agricultural trials, and crossover trials, as well as experiments on connected individuals in a social network.

Original language	English
Pages (from-to)	455-480
Journal	APPLIED STATISTICS
Volume	66
Issue number	3
Early online date	13 Aug 2016
DOIs	https://doi.org/10.1111/rssc.12170
Publication status	Published - Apr 2017

Keywords

design of experiments
social networks
optimal design

Access to Document

10.1111/rssc.12170

Optimal Design of Experiments_PARKER_Accepted30May2016_GREEN AAMAccepted author manuscript, 400 KBLicence: CC BY-NC

Cite this

@article{f1c6fc1d1dcd4803a61766ba800547d6,

title = "Optimal design of experiments on connected units with application to social networks",

abstract = "When experiments are performed on social networks, it is difficult to justify the usual assumption of treatment-unit additivity, due to the connections between actors in the network. We investigate how connections between experimental units affect the design of experiments on those experimental units. Specifically, where we have unstructured treatments, whose effects propagate according to a linear network effects model which we introduce, we show that optimal designs are no longer necessarily balanced; we further demonstrate how experiments which do not take a network effect into account can lead to much higher variance than necessary and/or a large bias. We show the use of this methodology in a very wide range of experiments in agricultural trials, and crossover trials, as well as experiments on connected individuals in a social network.",

keywords = "design of experiments, social networks, optimal design",

author = "Parker, {Ben M.} and Gilmour, {Steven G.} and Schormans, {John Alexander}",

year = "2017",

month = apr,

doi = "10.1111/rssc.12170",

language = "English",

volume = "66",

pages = "455--480",

journal = "APPLIED STATISTICS",

issn = "0035-9254",

publisher = "Wiley-Blackwell",

number = "3",

}

TY - JOUR

T1 - Optimal design of experiments on connected units with application to social networks

AU - Parker, Ben M.

AU - Gilmour, Steven G.

AU - Schormans, John Alexander

PY - 2017/4

Y1 - 2017/4

N2 - When experiments are performed on social networks, it is difficult to justify the usual assumption of treatment-unit additivity, due to the connections between actors in the network. We investigate how connections between experimental units affect the design of experiments on those experimental units. Specifically, where we have unstructured treatments, whose effects propagate according to a linear network effects model which we introduce, we show that optimal designs are no longer necessarily balanced; we further demonstrate how experiments which do not take a network effect into account can lead to much higher variance than necessary and/or a large bias. We show the use of this methodology in a very wide range of experiments in agricultural trials, and crossover trials, as well as experiments on connected individuals in a social network.

AB - When experiments are performed on social networks, it is difficult to justify the usual assumption of treatment-unit additivity, due to the connections between actors in the network. We investigate how connections between experimental units affect the design of experiments on those experimental units. Specifically, where we have unstructured treatments, whose effects propagate according to a linear network effects model which we introduce, we show that optimal designs are no longer necessarily balanced; we further demonstrate how experiments which do not take a network effect into account can lead to much higher variance than necessary and/or a large bias. We show the use of this methodology in a very wide range of experiments in agricultural trials, and crossover trials, as well as experiments on connected individuals in a social network.

KW - design of experiments

KW - social networks

KW - optimal design

U2 - 10.1111/rssc.12170

DO - 10.1111/rssc.12170

M3 - Article

SN - 0035-9254

VL - 66

SP - 455

EP - 480

JO - APPLIED STATISTICS

JF - APPLIED STATISTICS

IS - 3

ER -

Optimal design of experiments on connected units with application to social networks

Abstract

Keywords

Access to Document

Fingerprint

Cite this