Dataset Information

A new criterion for confounder selection.

ABSTRACT: We propose a new criterion for confounder selection when the underlying causal structure is unknown and only limited knowledge is available. We assume all covariates being considered are pretreatment variables and that for each covariate it is known (i) whether the covariate is a cause of treatment, and (ii) whether the covariate is a cause of the outcome. The causal relationships the covariates have with one another is assumed unknown. We propose that control be made for any covariate that is either a cause of treatment or of the outcome or both. We show that irrespective of the actual underlying causal structure, if any subset of the observed covariates suffices to control for confounding then the set of covariates chosen by our criterion will also suffice. We show that other, commonly used, criteria for confounding control do not have this property. We use formal theory concerning causal diagrams to prove our result but the application of the result does not rely on familiarity with causal diagrams. An investigator simply need ask, "Is the covariate a cause of the treatment?" and "Is the covariate a cause of the outcome?" If the answer to either question is "yes" then the covariate is included for confounder control. We discuss some additional covariate selection results that preserve unconfoundedness and that may be of interest when used with our criterion.

SUBMITTER: VanderWeele TJ

PROVIDER: S-EPMC3166439 | biostudies-literature | 2011 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A new criterion for confounder selection.

VanderWeele Tyler J TJ Shpitser Ilya I

Biometrics 20110531 4

We propose a new criterion for confounder selection when the underlying causal structure is unknown and only limited knowledge is available. We assume all covariates being considered are pretreatment variables and that for each covariate it is known (i) whether the covariate is a cause of treatment, and (ii) whether the covariate is a cause of the outcome. The causal relationships the covariates have with one another is assumed unknown. We propose that control be made for any covariate that is e ...[more]

PMID: 21627630

Similar Datasets

Project description:BackgroundResidual feed intake (RFI) is one measure of feed efficiency, which is usually obtained by multiple regression of feed intake (FI) on measures of production, body weight gain and tissue composition. If phenotypic regression is used, the resulting RFI is generally not genetically independent of production traits, whereas if RFI is computed using genetic regression coefficients, RFI and production traits are independent at the genetic level. The corresponding regression coefficients can be easily derived from the result of a multiple trait model that includes FI and production traits. However, this approach is difficult to apply in the case of multiple repeated measurements of FI and production traits. To overcome this difficulty, we used a structured antedependence approach to account for the longitudinality of the data with a phenotypic regression model or with different genetic and environmental regression coefficients [multi- structured antedependence model (SAD) regression model].ResultsAfter demonstrating the properties of RFI obtained by the multi-SAD regression model, we applied the two models to FI and production traits that were recorded for 2435 French Large White pigs over a 10-week period. Heritability estimates were moderate with both models. With the multi-SAD regression model, heritability estimates were quite stable over time, ranging from 0.14 ± 0.04 to 0.16 ± 0.05, while heritability estimates showed a U-shaped profile with the phenotypic regression model (ranging from 0.19 ± 0.06 to 0.28 ± 0.06). Estimates of genetic correlations between RFI at different time points followed the same pattern for the two models but higher estimates were obtained with the phenotypic regression model. Estimates of breeding values that can be used for selection were obtained by eigen-decomposition of the genetic covariance matrix. Correlations between these estimated breeding values obtained with the two models ranged from 0.66 to 0.83.ConclusionsThe multi-SAD model is preferred for the genetic analysis of longitudinal RFI because, compared to the phenotypic regression model, it provides RFI that are genetically independent of production traits at all time points. Furthermore, it can be applied even when production records are missing at certain time points.

Dataset Information

A new criterion for confounder selection.

Publications

A new criterion for confounder selection.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets