MixTwice

Implements large-scale hypothesis testing by variance mixing. It takes two statistics per testing unit – an estimated effect and its associated squared standard error – and fits a nonparametric, shape-constrained mixture separately on two latent parameters. It reports local false discovery rates (lfdr) and local false sign rates (lfsr). Manuscript describing algorithm of MixTwice: Zheng et al(2021) doi:10.1093/bioinformatics/btab162.