epiforecasts / Ebola-Expert-Interviews

Analysis of the results from interviews with experts regarding the spatial spread of Ebola in North-Kivu (2019/20)

Expert elicitation

Data and R code to support the expert elicitation analysis in Forecasting the geographic spread of Ebola Virus Disease in the Democratic Republic of the Congo during the 2018-2020 outbreak.

---

Study description

This study aims to forecast the geographic spread of the 2018-2020 Ebola Virus Disease outbreak in the DRC. The expert elicitation was conducted from December 2019 to March 2020, with a pilot study carried out in November 2019.

We describe the experts' responses and compare these to the cases reported.

---

Repository files

A description of each file and folder is provided below.

Sourced scripts

All script names starting by 00_ denote scrips that are not to be run alone but are sourced by other scripts.

• 00_data_1.R: R script that inputs expert forecasts for November 2019 (pilot study) and outputs a .csv file with the results, stored in Outputs.

• 00_data_2.R: R script that inputs expert forecasts for December 2019 and outputs 2 .csv files with the results: one including the additional health zones rated by the experts and one without, both stored in Outputs.

• 00_data_3.R: R script that inputs expert forecasts for January 2020 and outputs 2 .csv files with the results: one including the additional health zones rated by the experts and one without, both stored in Outputs.

• 00_data_4.R: R script that inputs expert forecasts for February 2020 and outputs 2 .csv files with the results: one including the additional health zones rated by the experts and one without, both stored in Outputs.

• 00_data_5.R: R script that inputs expert forecasts for March 2020 and outputs 2 .csv files with the results: one including the additional health zones rated by the experts and one without, both stored in Outputs.

• 00_data_cases.R: R script that reads in cases data.

• 00_cumulative_calc.R: R script that inputs the raw data for all forecasts and outputs the cumulative data, stored in Outputs.

• 00_plots.R: R script that inputs .csv files generated by 01_data.R and 02_data_cumulative.R and cases through 00_data_cases.R, merges the data sets and generates plots, stored in Plots.

Scripts to run

• 01_data.R: R script that sources scripts 00_data_1.R, 00_data_2.R, 00_data_3.R, 00_data_4.R, 00_data_5.Rand outputs .csv files, stored in Outputs.

• 02_data_cumulative.R: R script that inputs .csv files generated by 01_data.R, sources script 00_cumulative_calc.R and outputs cumulative forecasts, stored as .csv files in Outputs.

• 02_timeline.R: R script that inputs .csv file generated by 01_data.R and outputs a plot of the time line of expert elicitations, stored in Plots.

• 02_descriptive.R: R script that inputs .csv file generated by 01_data.R and generates descriptive analysis for the results text.

• 03_plots.R: R script that sources 00_plots.R and makes plots for each month, stored in Plots.

Folders

• Outputs: a folder containing the expert elicitation forecasts needed for analysis generated by the R scripts.

  • .csv files entitled results_month_year.csv include only forecasts for the HZs rated by all experts,
  • .csv files entitled results_month_year_additional_HZ.csv include only forecasts for the HZs rated by some experts (in response to the follow-up question "any additional HZs with >1 case with >5% prob?")
    • .csv files entitled results_month_year_cm.csv include only the HZs rated by all experts with cumulative probabilities .
  • .csv file entitled results_all.csv includes all .csv files entitled results_month_year.csv
  • .csv file entitled results_additional_HZ.csv includes all .csv files entitled results_month_year_additional_HZ.csv

• Plots: a folder to save the figures generated by the R scripts.

Project

• Expert-elicitation.Rproj: An RStudio project file, to avoid having to set your working directory to the folder on your computer.

---

Download the repository as a ZIP file using the green button Clone or download above, then open the .Rproj file in RStudio to begin.

The analysis was performed using R version 3.6.2 (2019-12-12).

For any issues with the code please contact Alicia Roselló.