Reproducing the results

Synthetic studies

We take linear model with Toeplitz design (in lm_tp folder) as an example. Linear model with equi-correlation design (in lm_eq folder), GLM with Toeplitz design (in glm_tp folder), and GLM with equi-correlation design (in glm_eq folder) are similar to this.

1. cv_hd_lm_tp.py contains the base program for running one replication of bootstrap CI algorithms. cv_hd_lm_tp.sub contains the scripts for loading the required module and running the .py file.
2. run_lm_tp.sh generates .py and .sub files for 1000 replications of bootstrap CI algorithms and submits the jobs to computing clusters.
3. postprocess_lm_tp.sh combines the output .csv files across 1000 replications.
4. hd_lm_tp_truerad_supp.py contains the base program for running one replication for simulating oracle widths. hd_lm_tp_truerad_supp.sub contains the scripts for loading the required module and running the .py file.
5. run_lm_tp.sh generates .py and .sub files for 500 replications for simulating oracle widths and submits the jobs to computing clusters. The results will be saved in .csv files.
6. postprocess_lm_tp_truerad.sh combines the output .csv files across 500 replications.
7. plots_lm_tp.py aggregates the results to compute empirical coverage probabilities, average widths, and oracle widths, and generates the plots.

Semi-synthetic studies

1. get_data.sh downloads and preprocesses the data files.
2. run_analysis_d200.py, run_analysis_d500.py, and run_analysis_d1000.py run the semi-synthetic analyses with dimensionality of 200, 500, and 1000, respectively, and store the results in "split" folder.
3. generate_plots.py generates the plots with using results from "split" folder.