Coronavirus reproduction rate Germany
The following reproduction number for covid 19 pandemic in Germany is estimated by using an analytical approach. This report is updated daily.
History of the reproduction number
reproduction number chart analysis Germany (7d)
Because the usually known number of cases with symptom onset during the past few days is still too low for extrapolating a meaningful nowcast, a cut is set at 4 days ago (onset of the disease). Additionally the 7-day reproduction number is moved to the left by one more day. This is just a bit of cosmetics to clean the difference between the time window of 7 days (center at 3.5) and the assumed generation time of about 4 days (center at 2.0).
The mapping for cases without a known onset of symptoms was adjusted somewhat more dynamically in order to anticipate changes in reporting behavior in the scenario of outbreaks (especially shorter time lags for tracking and testing).
During the latest outbreak in Guethersloh (North Rhine-Westphalia), the mass testing showed a faster diagnosis and reporting as usual by a comparatively large number of cases. For this reason, the time window for determining the mapping distribution function for cases without a known onset of illness was shifted to the right plus one day. In such a scenario, this adjustment should ensure that a changing diagnosis and reporting behavior is applied a little bit earlier to such cases. This dynamic should be also reflected in the estimated reproduction number.
Covid-19 cases (7 day moving average)
cases, 7d smoothed
dates: symptom onset
blue: cases, symptom onset known
green: cases, symptom onset by mapping
red: cases by nowcasting
History of estimates
Reprodution rate - recent estimates and changes (7d R)
dates: symptom onset
last adjustment 29 Jun, 2020 (symptom onset 24th)
The basic structure of this model is similar to the approach of the German Robert Koch Institute. The new Covid 19 cases, accumulated over 7 days, are compared with the time window one generation back (assumption: 4 days). The respective onset of illness is taken into account.
Cases without a known onset of disease are imputed using a mapping approach.
But there are some different procedures.
Unlike other professional approaches, a largely analytical way is preferred to estimate expected values, especially for nowcasting. Expectations are generally calculated outside of propability distributions, but of course using parameters that are derived from empirically observed characteristics. According to previous observations, this results in somewhat more stable estimates with fewer outliers. Significant changes are also imaged promptly.
The range of possible estimation errors and the associated confidence interval is derived from backtesting original nowcasting results in comparism with later reported cases with identical onset of illness.