Technology Category
Reasoning under Uncertainty: CausalityMachine Learning: Causal LearningCognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
Empirically assessing the plausibility of unconfoundedness in observational studies
🤖 AI Summary
In observational studies, unconfoundedness is empirically unverifiable—especially in the presence of unmeasured confounders. This paper proposes a model-free empirical test for conditional unconfoundedness that relies solely on temporal ordering among variables and assumptions about measurement error. The method assesses whether a given set of covariates suffices to satisfy the unconfoundedness condition without modeling confounding structure. Its core innovation lies in constructing a conditional association test based on the specificity of covariate predictions for exposure, enabling the first fully nonparametric, bias-assessment–driven evaluation of identifiability. We establish theoretical identification guarantees, demonstrate robust finite-sample statistical properties via simulation, and apply the method to the 1982 Pelotas Birth Cohort to delineate the identifiability boundary of the causal effect of childhood height on adult IQ.
Technology Category
Application Category
📝 Abstract
The possibility of unmeasured confounding is one of the main limitations for causal inference from observational studies. There are different methods for partially empirically assessing the plausibility of unconfoundedness. However, most currently available methods require (at least partial) assumptions about the confounding structure, which may be difficult to know in practice. In this paper we describe a simple strategy for empirically assessing the plausibility of conditional unconfoundedness (i.e., whether the candidate set of covariates suffices for confounding adjustment) which does not require any assumptions about the confounding structure, requiring instead assumptions related to temporal ordering between covariates, exposure and outcome (which can be guaranteed by design), measurement error and selection into the study. The proposed method essentially relies on testing the association between a subset of covariates (those associated with the exposure given all other covariates) and the outcome conditional on the remaining covariates and the exposure. We describe the assumptions underlying the method, provide proofs, use simulations to corroborate the theory and illustrate the method with an applied example assessing the causal effect of length-for-age measured in childhood and intelligence quotient measured in adulthood using data from the 1982 Pelotas (Brazil) birth cohort. We also discuss the implications of measurement error and some important limitations.
Problem
Research questions and friction points this paper is trying to address.
Assessing unconfoundedness plausibility without confounding structure assumptions
Testing covariate-outcome associations to validate adjustment sets
Evaluating conditional unconfoundedness in observational causal inference
Innovation
Methods, ideas, or system contributions that make the work stand out.
Assessing unconfoundedness without confounding structure assumptions
Testing covariate-exposure-outcome associations with temporal ordering
Using covariate subset analysis for conditional unconfoundedness evaluation
🔎 Similar Papers
No similar papers found.
