Metropolized Knockoff Sampling

Metropolized knockoff sampling is a flexible method for constructing exact model-X knockoffs using tools from the MCMC and graphical models literature. This page demonstrates this method in several illustrative settings using R and Python notebooks. Notably, a tutorial illustrating our method for arbitrary models can be found here.


All code is publically available through Github here.


“Metropolized Knockoff Sampling”

Stephen Bates, Emmanuel Candès, Lucas Janson, and Wenshuo Wang (2019). [PDF]


Stephen Bates and Wenshuo Wang are the authors of the software and maintain this webpage.