First 25 minutes of Maddow tonight is a must watch. Probably not online yet, but will be available as a podcast later tonight and msmbc will probably put up a vid. She discusses the craziness with herd immunity idea, the nutty cdc spokesperson, and the good dr Atlas, late of the Hoover Institution.
2 Likes
It’s insanity but sadly I feel numb at this point.
3 Likes
Yep. No vaccine for me for awhile.
My wife and I were sitting down to watch The Boys and delayed because of how intense that segment was.
The more accurate and direct response to me would have been:
“I believe that High SDI leads to lower cases and Low SDI leads to more cases. I also believe that testing data has been unreliable for several months. I don’t actually have strong evidence that any of this is true, but I still believe it. I’m not a scientist, am not a skilled data analyst, and I’m not interested in trying to prove to someone else that my beliefs are true.”
I woudn’t have any issue with that response - you don’t owe me (or anyone else) any kind of evidence. There’s no shame in not being a scientist or being a data analyst. I don’t know how to play the piano, and I’m not a very good swimmer. But it’s a fact that absolutely nothing you’ve shown represents any kind of objective, falsifiable evidence to support your beliefs. So I’m going to continue to dismiss the SDI discussions and graphs until/unless someone can point to evidence that they’re meaningful.
As for this:
I look forward to seeing your cite to me downplaying the pandemic.
I do remember being more optimistic than some other people on here, saying things like:
- Deaths in August were going to much closer to 2k than 3k (to someone who thought 3k was inevitable)
- Average daily deaths in August were unlikely to exceed 1.5k
- “My estimate for the peak is that the highest 7-day average deaths will be between 1,350-1,790 and will be somewhere around the 7 days from 8/2-8/8.”
Those were all in July and I feel pretty ok with them based on how things have gone since then:
But again, I look forward to seeing your evidence of me downplaying things.
7 Likes
And as for my current projection:
I think that in the next 3-4 weeks, we’ll see the 7-day average deaths increase, but I’d be surprised if they break 1,100-1,200 at any time before mid-October.
***To state the obvious, deaths of this magnitude are an extraordinarily bad thing.
5 Likes
For your part about downplaying it, I concede you’re right that it wasn’t as dramatic as I remembered. I seem to remember you saying we would have had a big reduction in deaths by a few weeks ago, but if not, I stand corrected and I apologize.
Do you treat your students like this? I am very interested in backing up my beliefs on this hypothesis, but you told me you are not interested in seeing that or showing me what you’re looking for. That makes me immediately think you want me to waste hours of my time backing up something that you already think is lol. This is a really rude approach to me, by the way, but whatever. I hope you treat your students better.
Why don’t you tell me what evidence you’re wanting to see like I asked you to do? I very likely am compiling it. In fact, here’s the first bit to describe how testing is bulls***. If you think it’s not, based on this, then I guess I really don’t get math and statistics at all.
Testing for the last 9 weeks starting with measurement period 7/15-7/21 showing peak tests, low tests, and most recent tests for 9/9-9/15:
| 9 Week Test Peak, Low, and Most Recent | Peak Tests in a Week Since 7/15 | Low Tests | 9/9-9/15 Tests |
|---|---|---|---|
| from covidtracking.com | |||
| U.S. Total | 5,639,799 | 4,543,275 | 4,543,275 |
| California | 887,847 | 660,222 | 660,222 |
| Texas | 444,894 | 204,015 | -91,584 |
| Florida | 414,507 | 141,007 | 151,974 |
| Arizona | 85,723 | 45,705 | 45,705 |
| Georgia | 231,073 | 117,264 | 117,264 |
| North Carolina | 204,543 | 163,472 | 182,463 |
| South Carolina | 82,130 | 48,295 | 60,586 |
| Tennessee | 181,198 | 133,240 | 164,438 |
| New York | 589,585 | 446,054 | 542,394 |
| Illinois | 391,814 | 261,678 | 332,284 |
| Alabama | 84,570 | 15,235 | 40,575 |
| Ohio | 209,986 | 151,958 | 209,986 |
| Louisiana | 173,553 | 78,529 | 123,604 |
| Virginia | 118,760 | 91,055 | 102,064 |
| Arkansas | 88,386 | 38,969 | 88,386 |
| Utah | 50,922 | 27,967 | 30,265 |
| Pennsylvania | 113,397 | 80,312 | 86,422 |
| Mississippi | 44,985 | 19,850 | 19,850 |
| Washington | 118,080 (actual high was bogus dump of weeks of tests) | 75,518 | 102,662 |
| Maryland | 104,090 | 62,184 | 62,184 |
| Minnesota | 103,129 | 45,997 | 45,997 |
| Oklahoma | 91,636 | 38,227 | 91,369 |
| Nevada | 52,064 | 21,067 | 22,484 |
| New Jersey | 211,698 | 96,724 | 141,565 |
| Iowa | 42,025 | 30,635 | 30,635 |
| Wisconsin | 97,385 | 52,515 | 63,627 |
| Indiana | 103,868 | 66,396 | 103,868 |
| Michigan | 216,650 | 192,139 | 207,537 |
| Missouri | 94,522 | 56,597 | 94,522 |
| Massachusetts | 149,770 | 80,967 | 103,277 |
| Colorado | 64,570 | 36,438 | 36,438 |
| Kentucky | 135,789 | 40,229 | 135,789 |
| Oregon | 42,393 | 26,565 | 26,565 |
| Nebraska | 28,375 | 14,994 | 28,375 |
| Kansas | 30,356 | 17,405 | 21,695 |
| New Mexico | 55,187 | 31,372 | 31,372 |
| Idaho | 20,787 | 10,932 | 10,932 |
| Connecticut | 109,362 | 71,076 | 103,884 |
| Delaware | 16,747 | 10,189 | 10,189 |
| South Dakota | 9,916 | 7,028 | 9,410 |
| Rhode Island | 21,626 | 10,306 | 18,602 |
| District of Columbia | 23,754 | 19,682 | 19,682 |
| West Virginia | 37,845 | 24,319 | 27,886 |
| New Hampshire | 12,964 | 8,753 | 9,495 |
| North Dakota | 12,753 | 8,516 | 8,516 |
| Maine | 33,971 | 16,661 | 33,783 |
| Wyoming | 8,276 | 3,589 | 8,276 |
| Montana | 30,995 | 11,100 | 20,588 |
| Alaska | 40,424 | 15,468 | 15,468 |
| Hawaii | 38,830 | 7,360 | 24,895 |
| Vermont | 16,368 | 4,810 | 4,810 |
Out of 9 measurement periods, do you think you can get a reliable picture of what’s going on based on just this simple chart? If you feel I’ve proven my point here about thinking testing results are very unreliable right now based on what I posted, tell me what you want to see next. Teach me, don’t condescend to me.
If you still feel those numbers are reliable and think they’re fine, give me a lay person explanation as to why because I absolutely do not see how.
The way that science works is you make a falsifiable prediction/assertion, then you see whether the data confirms or refutes that prediction/assertion.
If you believe that SDI predicts new cases in some particular way, then make that prediction and test it statistically.
If you believe that testing numbers are obviously wrong, present the decision rule that you’re relying on that tells you they’re wrong. Then show how the data points to the decision rule rejecting the data being reliable.
I have no interest in teaching you how to be an empirical researcher.
4 Likes
Cool, good talk.
Being a Trump supporter is a contributing health condition.
2 Likes
Where is the SDI raw data?
Fina-fucking-lly, masks are now mandatory in classrooms in the Czech Republic. Made absolutely no sense for them to be only on in hallways.
Guess going over 2,000 cases in a day will trigger some changes.
Looks like we made it through the week. Circumstances weren’t as bad as I thought. When the students on quarantine come back on the 25th, things will change big time because I bet at least one will have it and spread it through the school.
3 Likes
SDI alone surely matters, but there’s like many confounding variables at play.
GTFO with this reg nonsense. reghdfe is the new hotness.
From what I’ve seen, SDI is an important indicator but the threshold is difficult, if not impossible to predict. First you would need to adjust for population density, which is easy enough.
Then you’d have to adjust for acquired immunity’s impact on R0.
Then you’d have to adjust for mitigation compliance efforts, such as mask wearing. That’s pretty much impossible.
I’m also pretty sure you’d want to adjust for whether socializing was being done indoors or outdoors, and that’s an issue too.
Oh and circling back to population density, you also have to adjust for household size. Take the same apartment building with the same number of people and turn all the 2br apartments with 2 residents into two different apartments with studios and you’ve significantly impacted the area’s susceptibility to the virus.
So SDI is something worth analyzing, but as all these other variables are shifting, I am skeptical that it can be reliably predictive. Even if it’s predictive in the same area once or twice, mask usage going up or down can render the next prediction inaccurate.
An inaccurate shot was clearly taken at me over data I’ve compiled, but I’ll bite my tongue and keep it constructive, positive and focused on the merit of using SDI and it’s strengths/weaknesses and leave it at that.
1 Like
The issue is that SDI, as far as I know, doesn’t account for mask usage. Measuring mask usage in any scientific way is quite a challenge, and it can vary over time.
Yeah if you’re trying to see what impact closing bars or restaurants has, for example, it should be useful. But the impact that has on SDI in Philadelphia may not be the same as the impact that has on SDI in rural Oklahoma.
Over 40 percent of parents have already opted out of in-person classes, and that number is likely to grow, reflecting families’ deep frustration about the city’s reopening effort and skepticism about schools’ readiness.
2 Likes