Inferring COVID-19 spreading rates and potential change points for case number forecasts

Jonas Dehning; Johannes Zierenberg; Frank Paul Spitzner; Michael Wibral; Joao Pinheiro Neto; Michael Wilczek; Viola Priesemann

When comparing our inference based on three change points to the number of confirmed cases, we find them to largely match (Fig. 3 B,C) . Note that the tell-tale transient decrease in the data, which are expected independently of the weekend-related effects on sampling, are more evident in the raw number of newly confirmed cases (Fig. 3 B) than in the cumulative report (Fig. 3 C) . Such a transient decrease of new cases, before increasing again, originates from changing an exponential growth rate over small time-interval in the model. It is consistent with the observed temporary drop in newly confirmed cases and suggests a rapid implementation of mitigation measures by the public. However, we also observed a spread in the data points that was somewhat larger than expected by the model. We assign this to the fact that in the main model we did not incorporate an additional prior describing uncertainty and noise that is introduced by fluctuations in reporting (less reports on weekends, availability of test kits, etc.) -however, we verified that our results are consistent when we extend our main model to account for the week-related alterations ( Fig. S3 and S5 ). Given these and other additional sources of noise, we consider the match of model and data convincing.