lassoinf - Selective Inference for the Randomized Lasso

Confidence intervals and p-values for parameters chosen by the (randomized) lasso, valid after selection. Implements the polyhedral approach of Lee, Sun, Sun and Taylor (2016) <doi:10.1214/15-AOS1371>, the randomized response of Tian and Taylor (2018) <doi:10.1214/17-AOS1564> and the exact randomized inference of Panigrahi, Fry and Taylor (2024) <doi:10.1093/biomet/asae019>, with the truncated bivariate normal computations of Liu (2023) <doi:10.48550/arXiv.2308.10346>. Includes inference for user-specified estimands, screening of inactive variables and problems fitted by 'glmnet'. Computations use a 'C++' backend with matrix-free linear operators for wide designs.