[UPDATE: Added Germany + Canada]
[All data from Johns Hopkins CSSE]
Noticed a couple of interesting patterns when I put the growth rates for Confirmed and Deceased on the same chart, for a set of countries:
Let’s start by looking at China:
The graph shows the two growth factors, for Confirmed cases (green) and Deceased (red).
The interesting pattern to notice is that the two graphs tend to peak synchroniously, that is, if one goes up (or down), so does the other. On, or almost on the same day! I would have expected there to be a delay of perhaps a week or so, from when the confirmed factor peaks to when the deceased factor peaks, but no, that seems to occur (almost) simultaneously.
Next, let’s look at a regression plot to see if we can detect any relationship between the two variables:
Above we see that there is a clear relationship between the factors: when one grows, so does the other. And vice versa. Unfortunately, this graph does not tell us anything about the temporal aspect we observed above, that is, the remarkable syncronization in time, observed in the first chart.
Let’s look at Italy:
Also Italy exhibits this remarkable syncronization in time. Btw, what we also see here is that Italy now, over the past 2 weeks, seem to have managed to slow down the growth of both factors.
Regression for Italy:
Same pattern for Spain.
Also US exhibits the same pattern, however, there’s a difference here: while almost all the other countries I’ve looked at (I don’t present all of them here in this post) have the death growth factor generally ‘over’ the confirmed factor in the graphs, for US, until very recently, their growth rate for confirmed exceeded that of deceased.
Now, to something completely different, Sweden:
Now, the pattern of synchronization we’ve observed for the other countries has all but disappeared! Furhermore, the “gap” between the two growth factors is considerable larger for Sweden than for the others.
The regression plot also clearly reveals that “Sweden is different” – there’s a huge amount of uncertainty in the Bayesian regression for Sweden, in contrast to the other countries, where the set of regression lines were very tightly clustered. Furthermore, the (very vaque!) mean trend for Sweden is negative, while all the other countries have had positive mean slopes.
Finally, let’s look at Netherlands, a country that I just learned has a strategy somewhat similar to that of Sweden:
Also for Netherlands, the synchronization is less visible than for the countries. Furthermore, they seem also to have the death growth factor dominating, as is the case for Sweden.
Regression for Netherlands resemble that of most of the countries.
Germany: Germany is interesting, because they have a large number of Confirmed, but very few deaths registered. Still, the patterns hold.
For most of the countries I’ve looked at, there are two clear patterns:
- The positive regression relationship between the two growth factors
- A remarkable synchronization in time between the factors.
However, for Sweden neither of these patterns hold. Furthermore, Sweden has the largest “gap” between the two factors, possibly caused by very limited testing for Corona.
Netherlands also exhibit the lack of synchronization to some extent, and they also have a fairly large factor gap, but nowhere near as large as Sweden.
US is the only of these countries where the growth factor for confirmed consitently dominated that of deaths, until very recently.