Model-X knockoffs
A variable selection framework that achieves exact FDR control when the covariate distribution is known. Proposed by Barber and Candes in 2015, it constructs knockoff variables and compares importance statistics to perform selection.
Timeline (last 90 days)
Model-X knockoffs was proposed by Barber and Candès in 2015 and is a variable selection framework that achieves exact FDR control under the condition that the covariate distribution is known.
Model-X knockoffs does not depend on the model form of the outcome variable Y, and only requires that the distribution of the covariates X is known.
In practical applications, the distribution of the covariates X is often unknown or difficult to estimate accurately, and distribution estimation error directly undermines the theoretical guarantee of FDR control for Model-X knockoffs.
MAS and Model-X knockoffs are complementary: when the covariate distribution is known or can be estimated accurately, knockoffs may be the more efficient choice; when the distribution is unknown or difficult to model, MAS exhibits a model-agnostic advantage.
All Facts (4)
Model-X knockoffs was proposed by Barber and Candès in 2015 and is a variable selection framework that achieves exact FDR control under the condition that the covariate distribution is known.
50%UnverifiedModel-X knockoffs does not depend on the model form of the outcome variable Y, and only requires that the distribution of the covariates X is known.
50%UnverifiedIn practical applications, the distribution of the covariates X is often unknown or difficult to estimate accurately, and distribution estimation error directly undermines the theoretical guarantee of FDR control for Model-X knockoffs.
50%UnverifiedMAS and Model-X knockoffs are complementary: when the covariate distribution is known or can be estimated accurately, knockoffs may be the more efficient choice; when the distribution is unknown or difficult to model, MAS exhibits a model-agnostic advantage.
50%