ID | 63041 |
FullText URL | |
Author |
Suzuki, Etsuji
Department of Epidemiology, Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, Okayama University
Kaken ID
publons
researchmap
Tsuda, Toshihide
Department of Human Ecology, Graduate School of Environmental and Life Science, Okayama University
ORCID
Kaken ID
researchmap
Yamamoto, Eiji
Department of Information Science, Faculty of Informatics, Okayama University of Science
|
Abstract | Purpose: To show conditions of covariate balance for no confounding in the sufficient-cause model and discuss its relationship with exchangeability conditions.
Methods: We consider the link between the sufficient-cause model and the counterfactual model, emphasizing that the target population plays a key role when discussing these conditions. Furthermore, we incorporate sufficient causes within the directed acyclic graph framework. We propose to use each of the background factors in sufficient causes as representing a set of covariates of interest and discuss the presence of covariate balance by comparing joint distributions of the relevant background factors between the exposed and the unexposed groups. Results: We show conditions for partial covariate balance, covariate balance, and full covariate balance, each of which is stronger than partial exchangeability, exchangeability, and full exchangeability, respectively. This is consistent with the fact that the sufficient-cause model is a “finer” model than the counterfactual model. Conclusions: Covariate balance is a sufficient, but not a necessary, condition for no confounding irrespective of the target population. Although our conceptualization of covariate imbalance is closely related to the recently proposed counterfactual-based definition of a confounder, the concepts of covariate balance and confounder should be clearly distinguished. |
Keywords | bias
causality
confounding factors
epidemiologic methods
|
Note | © 2017 Elsevier Inc. This manuscript version is made available under the CC-BY-NC-ND 4.0 License.
http://creativecommons.org/licenses/by-nc-nd/4.0/.
This is the accepted manuscript version. The formal published version is available at [https://doi.org/10.1016/j.annepidem.2017.11.005] .
|
Published Date | 2018-1
|
Publication Title |
Annals of Epidemiology
|
Volume | volume28
|
Issue | issue1
|
Publisher | Elsevier BV
|
Start Page | 48
|
End Page | 53
|
ISSN | 1047-2797
|
Content Type |
Journal Article
|
language |
English
|
OAI-PMH Set |
岡山大学
|
Copyright Holders | © 2017 Elsevier Inc.
|
File Version | author
|
PubMed ID | |
DOI | |
Web of Science KeyUT | |
Related Url | isVersionOf https://doi.org/10.1016/j.annepidem.2017.11.005
|
License | http://creativecommons.org/licenses/by-nc-nd/4.0/
|
Citation | Etsuji Suzuki, Toshihide Tsuda, Eiji Yamamoto, Covariate balance for no confounding in the sufficient-cause model, Annals of Epidemiology, Volume 28, Issue 1, 2018, Pages 48-53.e2, ISSN 1047-2797, https://doi.org/10.1016/j.annepidem.2017.11.005.
|
Funder Name |
Japan Society for the Promotion of Science
|
助成番号 | JP26870383
JP15K08776
JP17K17898
|