Secondary Logo

Institutional members access full text with Ovid®

Share this article on:

Repair of Partly Misspecified Causal Diagrams

Oates, Chris J.a,b; Kasza, Jessicac; Simpson, Julie A.d; Forbes, Andrew B.c

Erratum

In the July 2017 issue of EPIDEMIOLOGY in the supplemental digital content to the article by Oates et al., “Repair of Partly Misspecified Causal Diagrams”, Proposition 3 was incorrect as stated. For the conclusion to hold in case (C2), an additional assumption is needed to ensure that the orientation phase of the PC algorithm is unaffected by expert misjudgment. The assumption, (A3), is the non-existence of triplets (i, j, k) such that (i)

and

, (ii)

and

, and (iii)

. All instances of (C2) in the paper satisfied (A3), thus the experimental results were unaffected. The corrected supplemental digital content can be viewed online (http://links.lww.com/EDE/B272). The authors are grateful to Marloes Maathuis and Emilija Perkovic for highlighting the problem with Proposition 3.

Epidemiology. 29(1):e10, January 2018.

doi: 10.1097/EDE.0000000000000659
Methods

Errors in causal diagrams elicited from experts can lead to the omission of important confounding variables from adjustment sets and render causal inferences invalid. In this report, a novel method is presented that repairs a misspecified causal diagram through the addition of edges. These edges are determined using a data-driven approach designed to provide improved statistical efficiency relative to de novo structure learning methods. Our main assumption is that the expert is “directionally informed,” meaning that “false” edges provided by the expert would not create cycles if added to the “true” causal diagram. The overall procedure is cast as a preprocessing technique that is agnostic to subsequent causal inferences. Results based on simulated data and data derived from an observational cohort illustrate the potential for data-assisted elicitation in epidemiologic applications. See video abstract at, http://links.lww.com/EDE/B208.

Supplemental Digital Content is available in the text.

From the aSchool of Mathematical and Physical Sciences, University of Technology Sydney, Sydney, NSW, Australia; bAustralian Research Council Centre of Excellence for Mathematical and Statistical Frontiers, Parkville, VIC, Australia; cDepartment of Epidemiology and Preventive Medicine, Monash University, Melbourne, VIC, Australia; and dCentre for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, VIC, Australia.

Submitted February 16, 2016; accepted March 21, 2017.

J.K. is supported by Australian National Health and Medical Research Council Centre of Excellence Grant 1035261, awarded to the Victorian Centre for Biostatistics (ViCBiostat). J.A.S. is funded by an Australian National Health and Medical Research Council (NHMRC) Senior Research Fellowship 1104975.

The authors Oates and Kasza equally contributed to this study.

The authors report no conflicts of interest.

Code for replication of the simulation study is available from oates.work/vetting.

Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (www.epidem.com).

Correspondence: Jessica Kasza, Department of Epidemiology and Preventive Medicine, Monash University, Alfred Centre, 99 Commercial Road, Melbourne, VIC 3004, Australia. E-mail: jessica.kasza@monash.edu.

Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.