Bayes factors for testing order constraints on variances of dependent outcomes

F. Böing-Messing*, J. Mulder

*Corresponding author for this work

Research output: Contribution to journalArticleScientificpeer-review

Abstract

In statistical practice, researchers commonly focus on patterns in the means of multiple dependent outcomes while treating variances as nuisance parameters. However, in fact, there are often substantive reasons to expect certain patterns in the variances of dependent outcomes as well. For example, in a repeated measures study, one may expect the variance of the outcome to increase over time if the difference between subjects becomes more pronounced over time because the subjects respond differently to a given treatment. Such expectations can be formulated as order constrained hypotheses on the variances of the dependent outcomes. Currently, however, no methods exist for testing such hypotheses in a direct manner. To fill this gap, we develop a Bayes factor for this challenging testing problem. Our Bayes factor is based on the multivariate normal distribution with an unstructured covariance matrix, which is often used to model dependent outcomes. Order constrained hypotheses can then be formulated on the variances on the diagonal of the covariance matrix. To compute Bayes factors between multiple order constrained hypotheses, a prior distribution needs to be specified under every hypothesis to be tested. Here, we use the encompassing prior approach in which priors under order constrained hypotheses are truncations of the prior under the unconstrained hypothesis. The resulting Bayes factor is fully automatic in the sense that no subjective priors need to be specified by the user.
Original languageEnglish
Number of pages10
JournalThe American Statistician
DOIs
Publication statusE-pub ahead of print - 2020

Fingerprint

Bayes Factor
Testing
Dependent
Covariance matrix
Testing Hypotheses
Repeated Measures
Multivariate Normal Distribution
Nuisance Parameter
Prior distribution
Truncation

Cite this

@article{8dd721a896e544ca83503bf128689de4,
title = "Bayes factors for testing order constraints on variances of dependent outcomes",
abstract = "In statistical practice, researchers commonly focus on patterns in the means of multiple dependent outcomes while treating variances as nuisance parameters. However, in fact, there are often substantive reasons to expect certain patterns in the variances of dependent outcomes as well. For example, in a repeated measures study, one may expect the variance of the outcome to increase over time if the difference between subjects becomes more pronounced over time because the subjects respond differently to a given treatment. Such expectations can be formulated as order constrained hypotheses on the variances of the dependent outcomes. Currently, however, no methods exist for testing such hypotheses in a direct manner. To fill this gap, we develop a Bayes factor for this challenging testing problem. Our Bayes factor is based on the multivariate normal distribution with an unstructured covariance matrix, which is often used to model dependent outcomes. Order constrained hypotheses can then be formulated on the variances on the diagonal of the covariance matrix. To compute Bayes factors between multiple order constrained hypotheses, a prior distribution needs to be specified under every hypothesis to be tested. Here, we use the encompassing prior approach in which priors under order constrained hypotheses are truncations of the prior under the unconstrained hypothesis. The resulting Bayes factor is fully automatic in the sense that no subjective priors need to be specified by the user.",
author = "F. B{\"o}ing-Messing and J. Mulder",
year = "2020",
doi = "10.1080/00031305.2020.1715257",
language = "English",
journal = "The American Statistician",
issn = "0003-1305",
publisher = "Taylor & Francis",

}

Bayes factors for testing order constraints on variances of dependent outcomes. / Böing-Messing, F.; Mulder, J.

In: The American Statistician, 2020.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Bayes factors for testing order constraints on variances of dependent outcomes

AU - Böing-Messing, F.

AU - Mulder, J.

PY - 2020

Y1 - 2020

N2 - In statistical practice, researchers commonly focus on patterns in the means of multiple dependent outcomes while treating variances as nuisance parameters. However, in fact, there are often substantive reasons to expect certain patterns in the variances of dependent outcomes as well. For example, in a repeated measures study, one may expect the variance of the outcome to increase over time if the difference between subjects becomes more pronounced over time because the subjects respond differently to a given treatment. Such expectations can be formulated as order constrained hypotheses on the variances of the dependent outcomes. Currently, however, no methods exist for testing such hypotheses in a direct manner. To fill this gap, we develop a Bayes factor for this challenging testing problem. Our Bayes factor is based on the multivariate normal distribution with an unstructured covariance matrix, which is often used to model dependent outcomes. Order constrained hypotheses can then be formulated on the variances on the diagonal of the covariance matrix. To compute Bayes factors between multiple order constrained hypotheses, a prior distribution needs to be specified under every hypothesis to be tested. Here, we use the encompassing prior approach in which priors under order constrained hypotheses are truncations of the prior under the unconstrained hypothesis. The resulting Bayes factor is fully automatic in the sense that no subjective priors need to be specified by the user.

AB - In statistical practice, researchers commonly focus on patterns in the means of multiple dependent outcomes while treating variances as nuisance parameters. However, in fact, there are often substantive reasons to expect certain patterns in the variances of dependent outcomes as well. For example, in a repeated measures study, one may expect the variance of the outcome to increase over time if the difference between subjects becomes more pronounced over time because the subjects respond differently to a given treatment. Such expectations can be formulated as order constrained hypotheses on the variances of the dependent outcomes. Currently, however, no methods exist for testing such hypotheses in a direct manner. To fill this gap, we develop a Bayes factor for this challenging testing problem. Our Bayes factor is based on the multivariate normal distribution with an unstructured covariance matrix, which is often used to model dependent outcomes. Order constrained hypotheses can then be formulated on the variances on the diagonal of the covariance matrix. To compute Bayes factors between multiple order constrained hypotheses, a prior distribution needs to be specified under every hypothesis to be tested. Here, we use the encompassing prior approach in which priors under order constrained hypotheses are truncations of the prior under the unconstrained hypothesis. The resulting Bayes factor is fully automatic in the sense that no subjective priors need to be specified by the user.

UR - https://app-eu.readspeaker.com/cgi-bin/rsent?customerid=10118&lang=en_us&readclass=rs_readArea&url=https%3A%2F%2Fwww.tandfonline.com%2Fdoi%2Ffull%2F10.1080%2F00031305.2020.1715257

U2 - 10.1080/00031305.2020.1715257

DO - 10.1080/00031305.2020.1715257

M3 - Article

JO - The American Statistician

JF - The American Statistician

SN - 0003-1305

ER -