Ensemble forecast performance by size

Introduction

Steps to run analysis

1. Run R/pull_and_save_case.R

2. Run R/pull_and_save_death.R

3. Run R/pull_and_save_hosp2.R

4. Run R/pull_and_save_fluhosp.R

5. Download multiyear forecast folders from: https://github.com/FluSightNetwork/cdc-flusight-ensemble/tree/first-papers/model-forecasts/component-models, and save the component-models/ folder into the raw-data/ folder

6. Run R/create-multiyear-lookuptable.R

7. If running analyses on TACC, need to make sure that all downloaded components are transferred to TACC system. Includes all truth data, lookup tables, and individual component forecasts.

8. Run tacc_scripts/create-scoring-task-script.R and tacc_scripts/multiyear-task-creation.R to get the task list for the supercomputer.

9. Update the corresponding slurm files for running: launcher/make-ensembles.slurm and launcher/make-multiyear.slurm

10. Run slurm script, which will call tacc_scripts/create-and-score-ensembles.R and score-multiyear-ensembles.R

11. Output on tacc is saved to respective folders in the work directory, these can be downloaded and saved locally into the processed-data/ folder.

12. Run the summarize scripts (summarize-recent-ensemble-scores.R and summarize-multiyear-ensemble-scores.R)

13. Run make-overall-forecast-plots.R

14. Run make-overall-forecast-plots-full.R

15. Run make-ensemble-building-plot.R

16. Run make-ts-performance-plots.R

17. Run make-horizon-performance-plots.R

18. Run make-density-rank-plots.R

19. Run make-prediction-interval-plots.R

20. Run construct-table1.R

21. Run pull-manuscript-stats.R to get some of the other statistics from short report

22. Run make-raw_forecast-plots.R

Contact

Contact Spencer Fox (sjfox@uga.edu) with any questions.