Can algorithms replace expert knowledge for causal inference? A case study on novice use of causal discovery.
With growing interest in causal inference and machine learning among epidemiologists, there is increasing discussion of causal discovery algorithms for guiding covariate selection. We present a case study of novice application of causal discovery tools and attempt to validate the results against a w...
| Published in: | American Journal of Epidemiology Vol. 194; no. 5; pp. 1399 - 1410 |
|---|---|
| Main Authors: | , |
| Format: | research tables/charts Journal Article |
| Published: |
Oxford University Press / USA
May2025
|
| Online Access: | View this record in EBSCOhost |
| fields | @attributes: recordID: 1 pdfLink: plink: https://search.ebscohost.com/login.aspx?direct=true&db=ccm&AN=185321822&site=ehost-live header: @attributes: shortDbName: ccm uiTerm: 185321822 longDbName: CINAHL Complete uiTag: AN controlInfo: bkinfo: dissinfo: jinfo: jid: 00029262 1X1 jtl: American Journal of Epidemiology issn: 00029262 maglogo: N pubinfo: dt: May2025 vid: 194 iid: 5 pid: 622 pub: Oxford University Press / USA artinfo: ui: 185321822 185321822 185321822 10.1093/aje/kwae338 185321822 ppf: 1399 ppct: 11 formats: tig: atl: Can algorithms replace expert knowledge for causal inference? A case study on novice use of causal discovery. aug: au: Gururaghavendran, Rajesh Murray, Eleanor J affil: Department of Epidemiology, Boston University School of Public Health, Boston, MA 02118, United States sug: subj: Machine Learning Algorithms Evaluation Machine Learning Algorithms Utilization Causality Medication Compliance Professional Knowledge Access to Information Human Secondary Analysis Case Studies Publication Bias Epidemiologists Software Causal Modeling Confounding Consensus Graphical User Interface Correlation Coefficient Chi Square Test Data Analysis Software Descriptive Statistics ab: With growing interest in causal inference and machine learning among epidemiologists, there is increasing discussion of causal discovery algorithms for guiding covariate selection. We present a case study of novice application of causal discovery tools and attempt to validate the results against a well-established causal relationship. As a case study, we attempted causal discovery of relationships relevant to the effect of adherence on mortality in the placebo arm of the Coronary Drug Project (CDP) data set. We used 4 algorithms available as existing software implementations and varied several model inputs. We identified 15 adjustment sets from 17 model parameterizations. When applied to a baseline covariate adjustment analysis, these 15 adjustment sets returned effect estimates with similar magnitude and direction of bias as prior published results. When using methods to control for time-varying confounding, there was generally more residual bias than compared to expert-selected adjustment sets. Although causal discovery algorithms can perform on par with expert knowledge, we do not recommend novice use of causal discovery without the input of experts in causal discovery. Expert support is recommended to aid in choosing the algorithm, selecting input parameters, assessing underlying assumptions, and finalizing selection of the adjustment variables. pubtype: Academic Journal doctype: research tables/charts Journal Article ougenre: Article language: English refInfo: holdings: @attributes: islocal: N |
|---|