Statistical properties of four effect-size measures for mediation models

Milica Miočević; Holly P. O’Rourke; David P. MacKinnon; Hendricks C. Brown

doi:10.3758/s13428-017-0870-1

Statistical properties of four effect-size measures for mediation models

Milica Miočević^*, Holly P. O’Rourke, David P. MacKinnon, Hendricks C. Brown

^*Corresponding author for this work

Psychiatry and Behavioral Sciences

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

149 Scopus citations

Abstract

This project examined the performance of classical and Bayesian estimators of four effect size measures for the indirect effect in a single-mediator model and a two-mediator model. Compared to the proportion and ratio mediation effect sizes, standardized mediation effect-size measures were relatively unbiased and efficient in the single-mediator model and the two-mediator model. Percentile and bias-corrected bootstrap interval estimates of ab/s_Y, and ab(s_X)/s_Y in the single-mediator model outperformed interval estimates of the proportion and ratio effect sizes in terms of power, Type I error rate, coverage, imbalance, and interval width. For the two-mediator model, standardized effect-size measures were superior to the proportion and ratio effect-size measures. Furthermore, it was found that Bayesian point and interval summaries of posterior distributions of standardized effect-size measures reduced excessive relative bias for certain parameter combinations. The standardized effect-size measures are the best effect-size measures for quantifying mediated effects.

Original language	English (US)
Pages (from-to)	285-301
Number of pages	17
Journal	Behavior Research Methods
Volume	50
Issue number	1
DOIs	https://doi.org/10.3758/s13428-017-0870-1
State	Published - Feb 1 2018

Funding

Acknowledgements This research was supported in part by National Institute on Mental Health, Grant No. R01MH040859 and National Institute on Drug Abuse, Grant No. R01DA009757. Part of this research was presented at the 2013 Society for Prevention Research Conference. This research was supported in part by National Institute on Mental Health, Grant No. R01MH040859 and National Institute on Drug Abuse, Grant No. R01DA009757. Part of this research was presented at the 2013 Society for Prevention Research Conference.

Keywords

Bayesian methods
Bias
Effect sizes
Efficiency
Interval estimates
Mediation

ASJC Scopus subject areas

Experimental and Cognitive Psychology
General Psychology
Developmental and Educational Psychology
Arts and Humanities (miscellaneous)
Psychology (miscellaneous)

Access to Document

10.3758/s13428-017-0870-1

Cite this

@article{748480b7c52548188be9b4aa908cf1bf,

title = "Statistical properties of four effect-size measures for mediation models",

abstract = "This project examined the performance of classical and Bayesian estimators of four effect size measures for the indirect effect in a single-mediator model and a two-mediator model. Compared to the proportion and ratio mediation effect sizes, standardized mediation effect-size measures were relatively unbiased and efficient in the single-mediator model and the two-mediator model. Percentile and bias-corrected bootstrap interval estimates of ab/sY, and ab(sX)/sY in the single-mediator model outperformed interval estimates of the proportion and ratio effect sizes in terms of power, Type I error rate, coverage, imbalance, and interval width. For the two-mediator model, standardized effect-size measures were superior to the proportion and ratio effect-size measures. Furthermore, it was found that Bayesian point and interval summaries of posterior distributions of standardized effect-size measures reduced excessive relative bias for certain parameter combinations. The standardized effect-size measures are the best effect-size measures for quantifying mediated effects.",

keywords = "Bayesian methods, Bias, Effect sizes, Efficiency, Interval estimates, Mediation",

author = "Milica Mio{\v c}evi{\'c} and O{\textquoteright}Rourke, {Holly P.} and MacKinnon, {David P.} and Brown, {Hendricks C.}",

note = "Funding Information: Acknowledgements This research was supported in part by National Institute on Mental Health, Grant No. R01MH040859 and National Institute on Drug Abuse, Grant No. R01DA009757. Part of this research was presented at the 2013 Society for Prevention Research Conference. Funding Information: This research was supported in part by National Institute on Mental Health, Grant No. R01MH040859 and National Institute on Drug Abuse, Grant No. R01DA009757. Part of this research was presented at the 2013 Society for Prevention Research Conference. Publisher Copyright: {\textcopyright} 2017, The Author(s).",

year = "2018",

month = feb,

day = "1",

doi = "10.3758/s13428-017-0870-1",

language = "English (US)",

volume = "50",

pages = "285--301",

journal = "Behavior Research Methods",

issn = "1554-351X",

publisher = "Springer Nature",

number = "1",

}

TY - JOUR

T1 - Statistical properties of four effect-size measures for mediation models

AU - Miočević, Milica

AU - O’Rourke, Holly P.

AU - MacKinnon, David P.

AU - Brown, Hendricks C.

N1 - Funding Information: Acknowledgements This research was supported in part by National Institute on Mental Health, Grant No. R01MH040859 and National Institute on Drug Abuse, Grant No. R01DA009757. Part of this research was presented at the 2013 Society for Prevention Research Conference. Funding Information: This research was supported in part by National Institute on Mental Health, Grant No. R01MH040859 and National Institute on Drug Abuse, Grant No. R01DA009757. Part of this research was presented at the 2013 Society for Prevention Research Conference. Publisher Copyright: © 2017, The Author(s).

PY - 2018/2/1

Y1 - 2018/2/1

N2 - This project examined the performance of classical and Bayesian estimators of four effect size measures for the indirect effect in a single-mediator model and a two-mediator model. Compared to the proportion and ratio mediation effect sizes, standardized mediation effect-size measures were relatively unbiased and efficient in the single-mediator model and the two-mediator model. Percentile and bias-corrected bootstrap interval estimates of ab/sY, and ab(sX)/sY in the single-mediator model outperformed interval estimates of the proportion and ratio effect sizes in terms of power, Type I error rate, coverage, imbalance, and interval width. For the two-mediator model, standardized effect-size measures were superior to the proportion and ratio effect-size measures. Furthermore, it was found that Bayesian point and interval summaries of posterior distributions of standardized effect-size measures reduced excessive relative bias for certain parameter combinations. The standardized effect-size measures are the best effect-size measures for quantifying mediated effects.

AB - This project examined the performance of classical and Bayesian estimators of four effect size measures for the indirect effect in a single-mediator model and a two-mediator model. Compared to the proportion and ratio mediation effect sizes, standardized mediation effect-size measures were relatively unbiased and efficient in the single-mediator model and the two-mediator model. Percentile and bias-corrected bootstrap interval estimates of ab/sY, and ab(sX)/sY in the single-mediator model outperformed interval estimates of the proportion and ratio effect sizes in terms of power, Type I error rate, coverage, imbalance, and interval width. For the two-mediator model, standardized effect-size measures were superior to the proportion and ratio effect-size measures. Furthermore, it was found that Bayesian point and interval summaries of posterior distributions of standardized effect-size measures reduced excessive relative bias for certain parameter combinations. The standardized effect-size measures are the best effect-size measures for quantifying mediated effects.

KW - Bayesian methods

KW - Bias

KW - Effect sizes

KW - Efficiency

KW - Interval estimates

KW - Mediation

UR - http://www.scopus.com/inward/record.url?scp=85016047465&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85016047465&partnerID=8YFLogxK

U2 - 10.3758/s13428-017-0870-1

DO - 10.3758/s13428-017-0870-1

M3 - Article

C2 - 28342072

AN - SCOPUS:85016047465

SN - 1554-351X

VL - 50

SP - 285

EP - 301

JO - Behavior Research Methods

JF - Behavior Research Methods

IS - 1

ER -

Statistical properties of four effect-size measures for mediation models

Abstract

Funding

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this