EUVIMEDOpdag EUVIMED
Open access · CC BY 4.0

Choosing Covariate Balancing Methods for Causal Inference: Practical Insights From a Simulation Study

Choosing Covariate Balancing Methods for Causal Inference: Praktische Einblicke aus einer Simulationsstudie

Etienne Peyrot iD, Raphaël Porcher iD, François Petit

Statistics in Medicine · 2026-07 · 45 (15-17)

DOI: 10.1002/sim.70672 ↗

Auf einen Blick

Eine Monte-Carlo-Simulation verglich mehrere Kovariaten-Balancing-Verfahren zur kausalen Inferenz. Die Leistung hing vom Schätzer und von der Szenariokomplexität ab.

Fragestellung

Wie verhalten sich IPTW, CBPS, CBPS-TLF, EB und KOM in verschiedenen Simulationsszenarien für kausale Inferenz?

Methode

Monte-Carlo-Simulationen in 36 Hauptszenarien mit variierender Stichprobengröße, Behandlungsprävalenz und Komplexitätsfaktor; zusätzlich PROBITsim als empirische Illustration.

Ergebnisse

Unter WLS waren IPTW und CBPS-TLF empfindlicher gegenüber Komplexität; standard- CBPS ähnelte oft IPTW. EB und KOM zeigten stabilere Punktschätzungen. DR reduzierte Unterschiede, die KI-Leistung blieb heterogen.

Einschränkungen

Die KIs, die Gewichtskonstruktion und Tuning berücksichtigen, bleiben ein offenes praktisches Problem.

Schlussfolgerung der Autor:innen

Die Studie ist als praktische Orientierung und nicht als Rangfolge der Methoden zu lesen; sie benennt Situationen, in denen Gewichtungsanalysen empfindlich auf Prävalenz, Overlap, Tuning und Varianzschätzung reagieren.

Kernaussagen

  • Leistung variierte je nach Schätzer und Szenario
  • EB und KOM zeigten stabilere Punktschätzungen
  • Konfidenzintervalle blieben methodisch problematisch

Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.

At a glance

Journal article in Statistics in Medicine, published online on 2026-07-08. Written by Etienne Peyrot and 2 co-authors. Institution: Université Paris Cité and Université Sorbonne Paris Nord, Inserm, INRAE, Center for Research in Epidemiology and StatisticS (CRESS), Paris, France. Major MEDLINE topics: Causality, Observational Studies as Topic. Cited 2 times so far (Crossref); 56 references. The full text is freely available (open access, CC BY 4.0). Funded by: Agence Nationale de la Recherche.

Auf einen BlickAbstractSubjectsBibliografiske oplysningerCiteRelated studies

Abstract (original language)

Background

Weighting methods are widely used for confounding adjustment in observational studies, but their finite‐sample behavior depends on implementation choices and empirical overlap. We compare IPTW, just‐ and over‐identified covariate balancing propensity score (CBPS), CBPS by tailored‐loss function (CBPS‐TLF), energy balancing (EB), and kernel optimal matching (KOM).

Methods

We conducted Monte Carlo simulations across 36 main scenarios varying sample size, treatment prevalence, and a complexity factor increasing confounding and reducing overlap. The main simulation considered a null constant treatment effect, with non‐null constant effects examined as sensitivity analyses. Average treatment effects and average treatment effects on the treated were estimated using weighted least squares (WLS) and doubly robust (DR) estimators. Inference followed published recommendations when feasible. An empirical illustration used PROBITsim .

Results

Performance depended on the estimator and scenario complexity. Under WLS, IPTW and CBPS‐TLF were more sensitive to complexity, while standard CBPS often behaved similarly to IPTW but with less deterioration in some high‐prevalence settings. EB and KOM showed more stable point‐estimation patterns across scenarios. DR estimation reduced differences between weighting methods when all confounders were included in the outcome model, although confidence‐interval performance remained heterogeneous. PROBITsim results were consistent with simulation patterns.

Conclusions

The study should be read as practical guidance rather than a ranking of methods. It identifies settings where weighting analyses become sensitive to prevalence, overlap, tuning, and variance estimation. Confidence intervals that account for weight construction and tuning remain an important open practical issue.

Subjects

MeSH terms (MEDLINE)

CausalityObservational Studies as TopicMonte Carlo MethodHumansComputer SimulationPropensity ScoreModels, StatisticalConfounding Factors, EpidemiologicSample SizeLeast-Squares Analysis

Author keywords

Monte Carlo simulationcausal inferenceinverse probability of treatment weightingobservational studytreatment effect estimation

Authors and institutions

  • Université Paris Cité and Université Sorbonne Paris Nord, Inserm, INRAE, Center for Research in Epidemiology and StatisticS (CRESS), Paris, France.
  • Centre d'Épidémiologie Clinique, Assistance Publique-Hôpitaux de Paris, Hôtel-Dieu, Paris, France.

ORCID profiles: Etienne Peyrot, Raphaël Porcher

Funding

  • Agence Nationale de la Recherche

Cite

APA

Peyrot, E., Porcher, R., & Petit, F. (2026). Choosing Covariate Balancing Methods for Causal Inference: Practical Insights From a Simulation Study. Statistics in Medicine, 45(15-17). https://doi.org/10.1002/sim.70672

Vancouver

Peyrot E, Porcher R, Petit F. Choosing Covariate Balancing Methods for Causal Inference: Practical Insights From a Simulation Study. Statistics in Medicine. 2026;45(15-17). doi:10.1002/sim.70672

BibTeX
@article{Peyrot2026,
  author = {Peyrot, Etienne and Porcher, Raphaël and Petit, François},
  title = {Choosing Covariate Balancing Methods for Causal Inference: Practical Insights From a Simulation Study},
  journal = {Statistics in Medicine},
  year = {2026},
  volume = {45},
  number = {15-17},
  doi = {10.1002/sim.70672}
}
Start for free →