This repository accompanies the manuscript Data-driven modeling reveals a universal dynamic underlying the COVID-19 pandemic under social distancing by Robert Marsland III and Pankaj Mehta. It contains four main items:
- A Jupyter notebook "COVID-19 Predictor" for plotting predictions alongside the latest fatality and case data. Click here to open the notebook, which is hosted on Binder for interactive use.
- Another Jupyter notebook "Coronavirus Figures" that generates all the figures in the manuscript.
- A snapshot of all the input data files (fatalities/cases, social policy metrics, and IHME predictions) required for generating the figures, from the time the analyses in the manuscript were carried out.
- The output data files generated by "Coronavirus Figures." The two files "predictions_cases_apr15.csv" and "predictions_deaths_apr15.csv" contain the best-fit values and 95 percent confidence bounds for the total final number of fatalities/cases
Nmax, the time of peak new fatalities/cases
th, and the infection timescale
sigmadefined in the text. Note that the confidence intervals are lower bounds on the true amount of uncertainty. These intervals are calculated under the assumption that the data is accurately described by our fitting function, with no changes in parameter values, and with all deviations from this function due to random multiplicative noise. The confidence intervals do not account for possible future changes in social policy, or for systematic deviations from the fitting function that may arise at late times.