Skip to content

Looking for what remains

  • data

What am I actually doing? What can I find out by looking at pictures based on COVID infection and death counts? Even the counts don’t match. And in the individual behavioral choices of people who consider themselves autonomous, little regularity can be expected.

First something about differences in counts. Below are two pictures of the situation in the US. They rest on different sources. The top picture shows the numbers as they are tracked and delivered by the Atlantic’s covidtracking project on a daily basis. The bottom picture shows the numbers as they are updated weekly and provided by the European Centre for Disease Prevention and Control.

The differences are subtle: the pictures are broadly alike. Local differences can mainly be attributed to the timescale: top in days and bottom in weeks. Above, a running 7-day average was used for both the numbers of deaths and infections (cases). Below there are straight lines between points that give the averages for that week per week. Two differences remain disturbing. The verticals show varying differences of 10-13 days. And the daily anchored image from around January 17 (above, of the Atlantic) appears to give data up to about 2 weeks later than below (of the EU).

The question is whether the difference of 37.279 between the total numbers of COVID deaths above and below could be explained by this time difference. At an average of 2,663 deaths per day, that difference could be explained in 14 days. In view of the numbers in the top graph it is likely that the average per day was higher: say around 3000. Then the bottom graph is 12 days behind. All in all, reason to give preference to the numbers of the Atlantic over those of the EU, although the use of the EU data seems stilljustifiable, especially when the focus is not mainly about the most recent data. Despite the differences, the two databases show sufficient similarities. These similarities indicate the legitimate expectation that the time series with observations of deaths and infections have a core of invariant knowledge, of knowledge that remains – albeit of knowledge that goes over a timeline, about observations during a certain period in a certain country. .

It is, of course, tempting to look further and consider whether such lasting knowledge can be derived from juxtaposing timelines. To get a first impression, I give the figures of the Atlantic for nine states in the US. The third column has three deep red republican states (Alabama, Kentucky, Oklahoma), the second column has three swing states (Georgia, Arizona, Florida) of which two fell to Biden and one to Trump in the 2020 election and the first column has three deep blue democratic states (California, New York, Massachusetts).

Maybe something to take a closer look at and analyze further.