NeurIPS 2020

Causal analysis of Covid-19 Spread in Germany

Meta Review

As noted by the reviewers, there are promising developments in this paper, and timely and interesting applied results. The technical quality can be further improved, however, given the extraordinary scientific challenges posed by the pandemic, I would argue that the paper would nevertheless be quite welcome in a dedicated COVID-19 track. Some suggestions for the authors, hopefully useful as they revise it: 1) while Germany had better case reporting then other countries, nevertheless, especially in the early epidemic, cases were severely underreported as compared to the true number of infections. Thus, the focus on case data, and on temporal ordering based on which state/district reported cases first, is a potential shortcoming. This could be addressed by focusing on deaths data, at least as a sensitivity analysis. 2) data on mobility is available from a number of different sources; this would provide a direct way to test the conclusions made by the SyPI about movement from one region to another. 3) the time period covered was until mid-May; updating the paper to include more recent data, and thus the impact of the relaxation of NPIs, would be very interesting. Also this would enable you to evaluate the predictions you made, by seeing whether the relaxation of any of the important NPIs you highlighted led to a resurgence in cases.