Here’s the officially reported coronavirus death toll through July 18. The raw data from Johns Hopkins is here.
8 thoughts on “Coronavirus Growth in Western Countries: July 18 Update”
Comments are closed.
Cats, charts, and politics
Cats, charts, and politics
Here’s the officially reported coronavirus death toll through July 18. The raw data from Johns Hopkins is here.
Comments are closed.
And still getting worse....
SD and WY infection rate are the only ones below 1, but that is a bit of an artifact. SD reports weekly, so you get a see-saw pattern in the averages, depending on how they're done. WY doesn't report on weekends, so you get a pattern their too.
SD has the lowest caseload, but I expect that to change the next time they report data. VT and NH just a bit higher, and then PA followed by the first peleton. It looks like the jump in caseloads for VT and NH might be leveling off--but need more data.
At the other end: AR in the lead, MO right behind, but falling back a little, and FL making a charge.
One point is that sd and wyo are not the ones still below 1.0. They were above 1.0 yesterday as every state was, per covidactnow.
Not sure if what you are saying re there is a reporting artifact is true or not. Covidactnow is pretty vague about how they calculate their infection rate estimate. For things like weekly averages, such as the case counts, weekly reporting will not really distort things as long as they report all the same day each week. It only delays the adjustment for a week but would never cause it to go the wrong direction.
Are you saying that covidactnow calculate the R just using a "zero" due to no reporting as if it was really a drop to zero cases? I tend to doubt that as it would be really dumb and would it not cause the R to drop to nearly on the 6th day of no reporting? And I see no distortion that big.
I am somewhat relying on covidactnow for an estimated R and assuming they are doing something that makes sense and presumably know more than I do. And have been using those Rs comparing across states, etc. t come to some tentative conclusions about delta and therefore some sort of optimal strategy ( which is now just let it spread).
But , as covidactnow does not explain how they do Rs, I have had to rely on faith.
The big issue to me for my analysis was the spread between places, how much the Rs differed between places with estimated immunity. And the spread just was too low as compared to what you expect with consensus assumptions re how things work, which made me adjust those base assumptions leading to different conclusions.
If the covidactnow method is garbage and especially if it has some error that incorrectly reduces the R range, then all my conclusions are garbage too.
Although covidactnow not adjusting for non reporting days would tend to go the other direction ( overestimating spread) it would also give me little confidence in their entire method.
I'm just reporting the values for today, hence "the only ones below". I did not mean to imply "still below"--sorry about that.
The R values given by CovidActNow are very much lagging indicators. If you look up each state individually, you'll see the current R value is a few days old. There is also a current recent estimate, but that value is not in the tables. The current recent estimate also has a huge error associated with it.
Their explanation is a bit vague:
"This one is a bit complicated. To calculate the infection growth, a mathematical model combines trends in daily new cases from approximately the last 14 days, with estimates for other variables, such as how many days on average occur between infection and transmission."
I would like to know if one of those values includes positive test rate.
Your values are a waste. You aren't adjusting for population size.
Also, just spot checking states and metro areas on covidactnow, I do see signs that the rate of change of R is leveled off and maybe now negative. Hard to clearly extrapolate to usa as a whole as i do not think they show a full usa R.
Now that is not nearly the same as the case count going down, but for a long time it seemed the Rs were not only over 1.0 but the Rs were increasing. So case counts were not only increasing exponentially, the rate of the increase in the increase was increasing. At least it seems that has stopped and maybe reversed..
Still maybe a long away from R going back below zero, but have to start somewhere. At some point, case counts would be high enough so that the increased immunity from addtl natural immunity will start to be noticeable.
In worst states, my guess is that the slow increase in total immunity per day is more attributable to addtl natural immunity than to addtl vaccinations. Seems like vaccinations are creeping along near .1% per day and might be getting to wear actual cases are approaching that too. Added together if you get addtl total immunity of .2% per day, you are starting to get there fast enough to be basically done by October or so.
Golack is a dumb moron. Most states have very low cases. New York isn't seeing any increased hospitalization. Besides thst, the number of vaccine false positives are ridiculous. They don't have Covid. Pure and simple.
Spades,
C'mon. You have some arguable points to make even if golack and I might not agree. I certainly would like to hear a reasoned argument and would be quite happy if you can argue a point and win ( the loser if a logical argument goes home made smarter, the winner does not).
But do you have to start with "golack is a moron". Why do that? First golack does not deserve that. Maybe some here are just as nasty on the other side and you can play these games with them. But golack, while it seems I usually disagree with him, stays generally civil and gives some good info.
I say this because it appears you do have some things to say and might be able to add to the conversation. But instead you mess it up with insults and nobody listens to what you say.
On your points , if I can understand them,
First I think you mean false positives in covid testing not false vaccine positives. Do not know what that would mean. If you mean false testing positives, I think there is some issue there that exaggerates cases and deaths to some extent but it is limited. And I break it into types. One type would be a "true false positive " where it shows the virus where no virus ever existed at all. But it seems this is quite rare. Why? Because this would show up in tests in places like New Zealand when they had wiped out the virus but still almost no positives. The upper bound for this is the lowest % of positive tests anyone gets, maybe .1% or lower. Not a big factor. The second is a testing error where the test magnifies a minuscule amount of virus or virus residue nowhere near infectious or contagious to get a positive result, largely because our cycle threshold was set to high. I expect this is what you are talking most about. It is a potential problem and has been discussed. But note a large part of that is the test detecting a stray reminding fragment of virus after someone was infected a month or two earlier and is now totally noninfectious ( a virus fragment cannot infect). But this is a bigger issue when case counts were much higher in prior months than current (like in may) than now when reverse. Back in may conceivable that these sort of false positives were greatly inflating the case count. Today not so much - just not enough cases in past months compared to today to have those virus fragments skew things a lot.
Third type is just plain human error where the test is fine and human messes up. And I suspect this is a bigger factor than most realize.
The problem in you pointing this out is that, to the extent it inflates the cases from the true number, that effect should have been higher in prior months than today. So maybe case counts are distorted high by testing errors but the estimated R is distorted too low.
And the real worry now is not the current case count so much . It is how fast they are increasing - the R.
You may have been accurate in arguing case counts were " very low " a few weeks or a month ago, even that was not really the point. But today I can still buy that they are "low" but "very low" is no longer a viable description.
And the issue is that, if they keep increasing at this rate, or even a slowly lower rate, they will not be "low" in any way for long.
Golack,
I did not think you really did even imply that but thought some might assume it.
And I have been wondering the same about whether they take into account positive test rate or the amount of testing ( same sort of adjustment).
You do seem to have delved into it a bit more than I have and I thought maybe you found out more than I did on what they were doing but appears not.
I would have felt better if they had some place explaining the details of the mathematical model. Even if it might be beyond me to understand, I could try and get some better idea. And even if I could not understand, just that it is available to see would give me more confidence.
Not that I think they are, but if they were deliberately trying to skew their results in some way ( more likely to exaggerate the problem) hiding what they are doing is what you expect. Sort of like a poll that reports results but does not disclose demographics or specific questions - I tend to dismiss those as garbage.
What did make me feel better is that, when I have tried to estimate R myself I think I get a higher R than they do..I suspect they are assuming a longer infection cycle than I am. I first noticed this back months ago when cases going down and their R smelled to me too high ( but below 1.0) . Any effect due to infection cycle length would reverse after you go above 1.0.