Looks like the U.S. will be at the top of this graph for a while.
In the next one there is not enough room for all the labels. The U.S. and UK are both near Sweden.
Where is the “this is super depressing” emoticon thingy or a broken heart?
:confounded: is pretty good 
yeah, and just feeling utterly deflated at the sheer unscientific stupidity and greed that we all are subject to.
lets hope we’re flattening the curve.
this graph from last week caught my eye, and makes me wonder if its really just that we’ve hit testing capacity.
and the cdc’s positivity rate continues to increase. ( more and more people are being tested who have the virus, than people who do not. )
The peak of a normal flu season in the U.S. would be less than 1 on this graph
Iceland might legitimately wonder if it’s overreacting
Whatever Ireland is doing, maybe it should do something else
it sounds like they switched this week to counting probable deaths instead of just confirmed. their numbers also seem maybe daily up to date.
the us reporting causes i guess depends per state. and the cdc numbers, according to what cnn said at least, lag by maybe a week or two. 
The Worldometers numbers seem to be pretty good, don’t know about artifacts on other sites
The stats keep doing this every Tuesday though
worldometers.info/coronavirus/country/us
I’ve noticed this weekly cyclical spike. There are all kinds of reporting artifacts happening. I think the 7-day rolling average is a better measure. Gives a smoother look and we can not only see if the average is changing, but we can test it for statistical significance. We can employ the rules for reading control charts:
I will do this as soon as I can get to it. Trying to finish an analysis right now… I am curious now.
OK, here it is. I used the data in the worldometers daily new cases to make this.
This is a 7-day running average with 95% confidence intervals on the daily 7-day mean, and +/-3SD control limits based only on April data.
By rules 2, 5 and 6 it looks like we are in a downward trend. However, look at the magnitude of the decrease. It is slow. It is around 2500 to 3000 fewer cases per two-week period. This means that at this same rate, it will take approximately 20 more weeks to go back to zero (-3000 cases every 14 days). 20 weeks from now is mid-September.
However, increased testing will inflate this average, as we are already starting to see a flattening of the decrease. Which is not to say we should not test. Absolutely we should test, because if we don’t then all of this data means absolutely nothing. We need to test, and get the most accurate reading.
I redid the chart with more data:
Control charts are “better” with the raw data (top). But these measurements are more suited for the 7-day rolling average, which bends the intent of statistical control a bit. They were designed for situations like Deming’s continual improvement scenarios of quality control of parts off a factory line. Not really designed for massive testing of virus patients. However, I see similarities.
We are measuring at least a couple things here. One - viral load in the population as it accumulates, and the quality of testing. That quality of testing (sensitivity, specificity) are a giant confounder here. If we test more, and the tests get more reliable over time, then it will appear that we never move much from the average, since the average shifts upward over time to accommodate the improved, increased measurements.
But even though that is clearly going on here, we can see important things. We can see that we have flattened the weekly incidence of coronavirus. We can see the increased testing in the last few days. We can see the amount of variation in the daily new cases and are ready to apply the rules each day as new data come in.
So, there is a lot here. The upshot? I believe we are flattening the curve. We are not yet on a clear downslope towards fewer cases. We are revealing the true prevalence of the virus. Later on, we will divide the data into everything pre 4/23 and post 4/23 when testing really kicked in.
we have flattened the curve of reported positive coronavirus cases.
that difference is critical. the cdc’s positivity rate ( lagging by about two weeks ) went up to 18.8% – that indicates we’re still way behind on testing. and likely means that more and more we’re just testing people with symptoms.
it could well mean ( no pun, i swear ) that our curve flattening is actually because of continuing test limitations. (edit: that our infection numbers are limited by our numbers of tests.) we should be seeing that positivity rate come down, i think the target is below 10%.
and to be really depressing… even if we we’re flattening the curve, georgia and florida are about to kick that trend right out the window.
For comparison, these are the official weekly death counts for the flu for the 2010s in the U.S.
How deadly was the flu in 2019? - Graphically Speaking
We might wonder whether these are the right numbers, given that the CDC counts “pneumonia” deaths separately, and these are reported counts rather than, you know, estimates from this and that.
But it seems reasonable to compare reported COVID deaths to reported flu deaths, with the understanding that both counts are underreported and undercounted and we can’t know everything.
As a sanity check we can see that this WaPo graph of the same information looks the same
Definitely all of the above. I think the operative word in all of this interpretation is not even a whole word. It is the ending: -ing. We are flattenING. It is not flat. We have not succeeded. There is no WIN. It could (and has) gone up again because of increased/better/more widespread testing. And we are nowhere near even the edge of the woods… all we are succeeding at is doing something to the incidence. Info on the ground corroborates this flattening - ICUs in most locales are not marching steadily towards being overwhelmed - the numbers of admissions and intubations is coming down.
yeah. this part is incredibly good news.
Today’s Tuesday spike is 12% down from two weeks ago
