@Donnie B.:
You say, about a certain graph,
This speaks for itself.
Respectfully...you sure about that?
I mean, are you confident that
that graph "speaks for itself" in the sense that what it seems to mean, at first glance, actually is what it means? And, that what it means to convey is even knowable-in-principle, given currently-available information?
I guess it depends on what one's first-glance conclusion is. But if the first thing I thought, on seeing that graph, was "
What the U.S. was doing sucked, in comparison to what
Spain and Germany were doing, from March 22 to April 11."
That was my first thought. Then I reconsidered.
That conclusion
may be true. But you couldn't possibly, even in principle,
draw such a conclusion from the data that graph is displaying. That graph is profoundly bad. If it should happen to prompt a true conclusion in someone's mind, it only does so the same way a stopped clock tells the "correct" time twice a day.
Think about it: It's "total cases" (reported to Johns Hopkins); i.e., confirmed by testing. But we know confirmed cases are some fraction of actual cases, and we know that doubling-rate-slowing happens for the following reasons:
- implementing social distancing
- diminishing percent-of-population infectable (by death, or recovery, or vaccination)
Well, the timing and degree of social distancing obviously varies from country-to-country. That's fair; presumably that's something the graph is trying to bring out. But, the percentage of population still infectable also varies (though we can't know by how much). That's a major variable that's not normalized.
And, we know that
artificial increases in "total cases" happens when you have more testing now vs, previously. (It reflects increasing awareness of reality, but it
looks like a failure of social distancing.)
And, we know that the countries listed had outbreaks that began at different times (why were those dates chosen?), with different numbers of initial vectors. That's setting aside other countries available for comparison. (Why these in particular? Where are Taiwan and South Korea? France? Israel? Canada? Australia? Why Spain if not also Greece and Portugal?) And of course test-availability has followed different timelines in each country as well.
And, we know that the percent-of-population still infectable is probably very different in Italy than, say, here. New York and Italy probably are closer to one another than anyplace in Kansas is to New York.
So here we have a graph with an orange line that probably ought to have been separate bars depicting five countries that, had they all responded to COVID-19 identically on a per-capita basis, and been equally successful, could
still have shown
wildly different outcomes on this graph (for the reasons stated above).
What would be required for that graph to become
validly informative, rather than
probably misleading is it is now? (Or, if it isn't misleading, it's only by accident?)
Well, first change it to all bars. (What're the slopes of the orange line
between countries supposed to convey, anyway? It's not like the U.S. gradually morphs into Spain on one side!)
Secondly, ensure that your cases are measured only through random testing of the general population using statistically-valid sample sets. You need to know the actual infection numbers, not something warped out-of-recognition by targeted testing.
Thirdly, don't arbitrarily use March 22 as your start date. Find out when, in each country, 5% of the population was infected. Use that date as the starting point
for that country, and then run your measurements forward.
Fourth, only make comparisons between comparable population-size regions with comparable mixes of urban and rural living. Don't compare France to the whole U.S.A.; compare it to the state of Georgia, perhaps.
Finally,
maybe make the vertical axis a percentage of population axis, not an absolute numbers axis.
Maybe. None of my other four requirements should be controversial, but this one could be. On the one hand, we want to save
lives, not "percentages of populations." On the other hand, the entire population of Lichtenstein could be infected and they'd still have such a low caseload they'd look like heroes.
Anyway, without those kinds of changes, that graph is, I think, non-informative. The U.S. response from March 22 to April 11 may have sucked; but that graph is no good reason to think so.
What reasons there are must be found elsewhere.