/Forecasting-Tournament

Data repository of the results for the COVID-19 Tournament of the Behavioral and Social Science Forecasting Collaborative

Primary LanguageJupyter Notebook

This repository contains the data behind COVID-19 Tournament of the Forecasting Collaborative.

Data

  • dat_for_analyses.csv contains a harmonized file used for main analyses of forecasting accuracy.
  • dat_long.csv contains a file with predictions of forecasting teams in a long format, used for plotting estimates of each team.
  • sim folder contains R files used to simulate naive benchmarks
  • historical_data.csv contains historical data provided to participants for domains they chose to make forecasts for, as well as ground truth scores for each domain, along with information on a key conditional factor - COVID-19 infections and deaths in the US.
  • dat_for_analyses.csv contains a harmonized file used for main analyses of forecasting accuracy.
  • wave1.scores.csv contains the estimates in the First Tournament (May 2020) for each month per domain, ranked in terms of inaccuracy (MASE scores) of Forecasting Teams.
  • wave2.scores.csv contains the estimates in the Second Tournament (Nov 2020) for each month per domain, ranked in terms of inaccuracy (MASE scores) of Forecasting Teams.
  • top.t1.csv.csv contains the estimates of the best performing teams in each domain in the First Tournament (May 2020) per month, ranked in terms of inaccuracy (MASE scores) of Forecasting Teams, along with the forecasting approach, and team characteristics.
  • top.t2.csv.csv contains the estimates of the best performing teams in each domain in the Second Tournament (Nov 2020) per month, ranked in terms of inaccuracy (MASE scores) of Forecasting Teams, along with the forecasting approach, and team characteristics.
  • Columns reference.xlsx contains the legend for all variable names in the dataset.

Method

Analyses