This is a collection of scripts in R and Python that sample exact knockoffs for graphical models via the method of Metropolized knockoff sampling (Metro). Scripts are grouped by the distribution of the original features. They reproduce the results in Bates, Candès, Janson and Wang (2020), including knockoff sampling for discrete Markov chains, continuous Markov chains (Gaussian, heavy-tailed, and asymmetric), Ising/Gibbs models on grids and Potts model on a graph inspired by a real problem in protein contact prediction. The script for the Potts model implements a general knockoff sampler that works for any general graphical model.

A tutorial can be found here. Correspondence should be addressed to Stephen Bates and Wenshuo Wang.

Stephen Bates, Emmanuel Candès, Lucas Janson and Wenshuo Wang. (2020). Metropolized Knockoff Sampling. Journal of the American Statistical Association (to appear). [pdf] [arXiv] [journal]