2018/12/12 by Lars Holden, Kristoffer H. Hellton, Holden, Lars +1
Mathematics · #62F03 #62J05 #FOS: Computer and information sciences #Methodology (stat.ME) #Statistical Methods and Bayesian Inference #Statistical Methods and Inference #Statistical Methods in Clinical Trials
paper · pdf · doi:10.48550/arxiv.1812.04928
openalex publication_date 2018/12/12 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Model-free knockoffs is a recently proposed technique for identifying covariates that is likely to have an effect on a response variable. The method is an efficient method to control the false discovery rate in hypothesis tests for separate covariates. This paper presents a generalisation of the technique using multiple sets of model-free knockoffs. This is formulated as an open question in Candes et al. [4]. With multiple knockoffs, we are able to reduce the randomness in the knockoffs, making the result stronger. Since we use the same structure for generating all the knockoffs, the computational resources is far smaller than proportional with the number of knockoffs. We prove a bound on the asymptotic false discovery rate when the number of sets increases that is better then the published bounds for one set.