This is an update as of April 2nd, 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.
Simple exponential fit is not included, as most of the regions are already experiencing mature outbreaks.
According to the fit, Italy and Spain have passed the peak, while Louisiana and Netherlands are a couple of days away from reaching the peak.
In order to make better understanding of the data available and possible correlations, some external data is gathered and correlated with the fitted parameters. The gathered data are as follows:
- Mean temperature for the month of March 2020. We want to analyze the claim that warmer weather has a dampening effect on the outbreak.
- Mean dewpoint for the month of March 2020. We want to analyze whether humid air has an effect on the outcome of the disease.
- Population density in the largest metro area of the state.
- Number of riders per week in the mass transit system of the area. We want to see if the mass transit system has an impact on the spreading of the disease.
- The population size of the largest metropolitan area in the state.
Correlation between those external data and the parameters of the fit
What we can see from the above table is that as expected the temperature and dew point give similar results. The dewpoint is slightly better correlated, but given how noisy the data is, it’s within any reasonable margin.
The number of riders and the size give similar results and they are both correlated with the outbreak size. That is expected since the number of riders captures the size of the area too. Interestingly, while the size and the population density are not correlated with the speed of the outbreak, the number of mass transit rides does show a small correlation.
Lastly, we can see that the population density is correlated with the size of the outbreak, even if it doesn’t capture directly the information about the size of the area. One speculation is that the speed of infection is much larger than the speed of mortality in dense urban areas and what we see for speed of outbreak(b parameter) for New York, Spain, or Michigan (0.09) are just the death rate lagged response to a much faster infection pulse earlier in time.
Three scatter plots are shown for these external factors correlated with fitted parameter for each of the US states.
Error function Fit
Correlation of external factors with fitted parameter for each of the US states
Your feedback is important to us. Please send it to feedback@rotellacapital.com