Network autocorrelation modeling

A Bayes factor approach for testing (multiple) precise and interval hypotheses

D. Dittrich*, R.T.A.J. Leenders, J. Mulder

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

52 Downloads (Pure)

Abstract

Currently available (classical) testing procedures for the network autocorrelation can only be used for falsifying a precise null hypothesis of no network effect. Classical methods can be neither used for quantifying evidence for the null nor for testing multiple hypotheses simultaneously. This article presents flexible Bayes factor testing procedures that do not have these limitations. We propose Bayes factors based on an empirical and a uniform prior for the network effect, respectively, first. Next, we develop a fractional Bayes factor where a default prior is automatically constructed. Simulation results suggest that the first two Bayes factors show superior performance and are the Bayes factors we recommend. We apply the recommended Bayes factors to three data sets from the literature and compare the results to those coming from classical analyses using p values. R code for efficient computation of the Bayes factors is provided.
Original languageEnglish
Pages (from-to)642-676
JournalSociological Methods and Research
Volume48
Issue number3
DOIs
Publication statusPublished - 2019

Fingerprint

testing procedure
simulation
performance
evidence
literature

Keywords

  • network autocorrelation model
  • hypothesis testing
  • Bayes factor
  • informative prior
  • fractional Bayes factor

Cite this

@article{461af3052fe54d7fbaeb8ba5f724e1c1,
title = "Network autocorrelation modeling: A Bayes factor approach for testing (multiple) precise and interval hypotheses",
abstract = "Currently available (classical) testing procedures for the network autocorrelation can only be used for falsifying a precise null hypothesis of no network effect. Classical methods can be neither used for quantifying evidence for the null nor for testing multiple hypotheses simultaneously. This article presents flexible Bayes factor testing procedures that do not have these limitations. We propose Bayes factors based on an empirical and a uniform prior for the network effect, respectively, first. Next, we develop a fractional Bayes factor where a default prior is automatically constructed. Simulation results suggest that the first two Bayes factors show superior performance and are the Bayes factors we recommend. We apply the recommended Bayes factors to three data sets from the literature and compare the results to those coming from classical analyses using p values. R code for efficient computation of the Bayes factors is provided.",
keywords = "network autocorrelation model, hypothesis testing, Bayes factor, informative prior, fractional Bayes factor",
author = "D. Dittrich and R.T.A.J. Leenders and J. Mulder",
year = "2019",
doi = "10.1177/0049124117729712",
language = "English",
volume = "48",
pages = "642--676",
journal = "Sociological Methods and Research",
issn = "0049-1241",
publisher = "SAGE Publications Inc.",
number = "3",

}

Network autocorrelation modeling : A Bayes factor approach for testing (multiple) precise and interval hypotheses. / Dittrich, D.; Leenders, R.T.A.J.; Mulder, J.

In: Sociological Methods and Research, Vol. 48, No. 3, 2019, p. 642-676.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Network autocorrelation modeling

T2 - A Bayes factor approach for testing (multiple) precise and interval hypotheses

AU - Dittrich, D.

AU - Leenders, R.T.A.J.

AU - Mulder, J.

PY - 2019

Y1 - 2019

N2 - Currently available (classical) testing procedures for the network autocorrelation can only be used for falsifying a precise null hypothesis of no network effect. Classical methods can be neither used for quantifying evidence for the null nor for testing multiple hypotheses simultaneously. This article presents flexible Bayes factor testing procedures that do not have these limitations. We propose Bayes factors based on an empirical and a uniform prior for the network effect, respectively, first. Next, we develop a fractional Bayes factor where a default prior is automatically constructed. Simulation results suggest that the first two Bayes factors show superior performance and are the Bayes factors we recommend. We apply the recommended Bayes factors to three data sets from the literature and compare the results to those coming from classical analyses using p values. R code for efficient computation of the Bayes factors is provided.

AB - Currently available (classical) testing procedures for the network autocorrelation can only be used for falsifying a precise null hypothesis of no network effect. Classical methods can be neither used for quantifying evidence for the null nor for testing multiple hypotheses simultaneously. This article presents flexible Bayes factor testing procedures that do not have these limitations. We propose Bayes factors based on an empirical and a uniform prior for the network effect, respectively, first. Next, we develop a fractional Bayes factor where a default prior is automatically constructed. Simulation results suggest that the first two Bayes factors show superior performance and are the Bayes factors we recommend. We apply the recommended Bayes factors to three data sets from the literature and compare the results to those coming from classical analyses using p values. R code for efficient computation of the Bayes factors is provided.

KW - network autocorrelation model

KW - hypothesis testing

KW - Bayes factor

KW - informative prior

KW - fractional Bayes factor

U2 - 10.1177/0049124117729712

DO - 10.1177/0049124117729712

M3 - Article

VL - 48

SP - 642

EP - 676

JO - Sociological Methods and Research

JF - Sociological Methods and Research

SN - 0049-1241

IS - 3

ER -