This repo hosts the code used to generate figures in the cKnockoff paper.

R/: host R code for the numerical experiments

data/: store the numerical results

figs/: R/plot.R will read the numerical results in data/ and save the figures in figs/

jobs/: Slurm job files to be submitted to cluster for large scale simulation

log/: store the Slurm logs.

Required packages:

• dbh: devtools::install_github("lihualei71/dbh")
• knockoff
• cknockoff: devtools::install_github("yixiangluo/cknockoff")
• glmnet
• KernSmooth
• here
• abind
• foreach
• doParallel

1. Create a my_expr_setting.R file in R/settings (see existed examples), in which specify the experiment settings.
2. Run Rscript R/do_expr.R my_expr_setting.
3. The real time progress are shown in data/temp/progress-my_expr_setting.txt, the results are stored in data/my_expr_setting.RData, and the produced figure is figs/simu-my_expr_setting.pdf.

1. Echo the names of the experiments (e.g. my_expr_setting) in jobs/expr_names.sh
2. Specify the Slurm arguments in jobs/expr_template.sh according your cluster settings. e.g. change the notification email at line 11. But please do NOT add/delete lines or change the order of lines, since this is a template file the content will be read and modified by R/cluster/set_expr.R according to which line the content is in.
3. Run jobs/set_exprs.sh (e.g. sbatch jobs/set_exprs.sh). This will prepare the shared data and structure used by each task in cluster computing.
4. Submit the jobs to the cluster by sh jobs/do_exprs.sh.
5. Check the progress at sh R/cluster/show_progress.R.
6. After all the experiments are done, run sh jobs/post_process.sh. This will aggregate the numerical results from all tasks and draw figures.
7. If there is an error, check the log in log/.

Modify the experiment settings in R/illustrative/scale_m_expr.R or R/illustrative/scale_n_expr.R and run them.

Run the files in R/HIV_drug_expr sequentially: HIV_preprocess.R, HIV_expr.R, postprocess_HIV.R, and plot_HIV_expr.R.