This is an update as of April 4th, 2020.
We are analyzing the existing available data on daily deaths caused by the SARS-CoV2 virus and use that in conjunction with certain simple models to predict the evolution of the disease in certain geographical areas. You can read the full introduction here.
Sharp increase in deaths in France puts it on a different curve.
Italy and Spain already passed the peak, for Netherlands are about to peak now. New York, Michigan, Louisiana and Colorado should peak within the next few days.
Another fit is included in today’s analysis that assumes that the daily distribution of casualties is log-normal, rather than just normal. The parameters of the log-normal fit are below.
The AICc coefficient for the log-normal distribution is slightly higher than the normal one. The bad news is that the log-normal distribution projects a higher number of total casualties (size) than the normal distribution, as the die-off in deaths is slower than the ramp-up. The SIR model also suggests a distribution that can be approximated by a log-normal. We will include skewed-normal distribution with four parameters as well in the future.
Correlation between those external data and the parameters of the fit
The speed of the outbreak correlates positively most with the mass transit rides and negatively with the Dew Point. We are using the average value of dew point for the month of March, because the dew point captures the absolute humidity (the total amount of water vapor in the atmosphere) and is the most commonly used and reported proxy for it.
Two scatter plots are showing these correlations.
The next hot zone of the outbreak seems to be Indiana. It will be added on Monday to the analysis and there will be another external parameter to analyze.
Error Function Fit
Log-normal Parameters
Your feedback is important to us. Please send it to feedback@rotellacapital.com