This repository serves as companion to the following paper:
N. Ignatiadis and W. Huber (2021), Covariate powered cross-weighted multiple testing, Journal of the Royal Statistical Society: Series B (JRSS-B)
The paper is also available on [arXiv].
All numerical results and figures in the aforementioned paper are made third-party reproducible herein.
We note that the Bioconductor package IHW provides a user-friendly implementation of the IHW-BH/Storey procedures with conditional distributions estimated with the Grenander estimator.
First, below we provide links to pre-rendered vignettes that reproduce figures.
- hQTL data analysis example (Figs. 1 and 7) [Vignette]
- Simulations on grouped multiple testing (Figs. 2 and 3) [Vignette]
- Simulations on multiple testing with continuous covariates (Figs. 4 and 5) [Vignette]
- Simulations on simultaneous two-sample testing (Fig. 6) [Vignette]
This directory contains IHWStatsPaper, a R package wrapping/implementing the different methods compared, the simulation functions, as well as the benchmarking code. It can be installed as follows.
devtools::install_github("Huber-group-EMBL/covariate-powered-cross-weighted-multiple-testing",
subdir="IHWStatsPaper")The simulations in the paper were implemented based on the following commit of the package.
R scripts that run the simulations (see in folder for details).
Simulation results from scripts in the scripts directory that have been pre-computed/cached.
R markdown documents that produce the pre-rendered vignettes above upon being compiled (using files saved in the precomputed_results folder). The only exception is the vignette for the hQTL data analysis example, which we include as part of the Bioconductor package IHWpaper.