Abstract
Controlling for confounding bias is crucial in causal inference. Distinct methods are currently employed to mitigate the effects of confounding bias. Each requires the introduction of a set of covariates, which remains difficult to choose, especially regarding the different methods. We conduct a simulation study to compare the relative performance results obtained by using four different sets of covariates (those causing the outcome, those causing the treatment allocation, those causing both the outcome and the treatment allocation, and all the covariates) and four methods: g-computation, inverse probability of treatment weighting, full matching and targeted maximum likelihood estimator. Our simulations are in the context of a binary treatment, a binary outcome and baseline confounders. The simulations suggest that considering all the covariates causing the outcome led to the lowest bias and variance, particularly for g-computation. The consideration of all the covariates did not decrease the bias but significantly reduced the power. We apply these methods to two real-world examples that have clinical relevance, thereby illustrating the real-world importance of using these methods. We propose an R package RISCA to encourage the use of g-computation in causal inference.
Highlights
Controlling for confounding bias is crucial in causal inference
Amongst the several existing doubly robust estimators (DRE), we focused on the targeted maximum likelihood estimator (TMLE)[24], for which estimators of application to multiple sclerosis (ATE) and ATT have been proposed[47]
TMLE experienced some difficulty in converging for ATT estimation in the medium-sized sample (n = 300)
Summary
Controlling for confounding bias is crucial in causal inference. Distinct methods are currently employed to mitigate the effects of confounding bias. Observational studies have the advantage of limiting the issue of external validity, but treated and untreated patients are often non-comparable, leading to a high risk of confounding bias To reduce such confounding bias, the vast majority of observational studies have been based on multivariable models (mainly linear, logistic, or Cox models), allowing for the direct estimation of conditional (subject-specific) effects, i.e., the average effect across sub-populations of subjects who share the same characteristics. Some authors[23,24] have proposed combinations of GC and PS to improve the estimation of the marginal causal effect These methods are known as doubly robust estimators (DRE) because they require the specification of both the outcome (for GC) and treatment allocation (for PS) mechanisms to minimise the impact of model misspecification. These estimators are consistent as long as either the outcome model or the treatment model is estimated correctly[25]
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have