5 ms·

Here's a useful twitter thread that talks about mistakes these doctors made: https://twitter.com/CT_Bergstrom/status/1254481543759683584?s=20 https://twitter.co

by DanBC 7y ago

Here's a useful twitter thread that talks about mistakes these doctors made: https://twitter.com/CT_Bergstrom/status/1254481543759683584?s=20 https://twitter.com/CT_Bergstrom/status/1254481543759683584?... > What they did was simple: they looked at the fraction of patients who tested positive for #COVID19 at the clinics they own. They found 340 out of 5213 tests were postive, about 6.6% > Then they assume the same fraction of the whole population are infected.

lurquer 7y ago
That was an obvious error in their reasoning. Not surprising, though. Idiotic (and, yes, such an error IS idiotic) fumbling with statistics is displayed on a daily basis. It doesn't matter what 'side' you are on: idiots abound. Doomer fallacies: Charts showing exponential growth of positives, without taking into account the exponential growth of tests... Flu-Bro fallacies: Antibody surveys that don't take into account false positives and make no mention of the expected number of positives in a population (both needed to make ANY sense of the results.) Doomer fallacies: No distinction between dying 'with' versus dying 'from'. Flu-Bro fallacies: Relying on death counts from the Chinese Communist Party. Doomer fallacies: Comparing 'positive' test rates between populations and countries with utterly different criteria for administering tests. etc. One thing you can count on, though: You will usually get a downvote on Hacker News if you deviate from the mainstream.
- alexpetralia 7y ago
  Yes I have historically found many doctors to not understand basic statistics, which has somewhat surprised me.
- chimprich 7y ago
  > One thing you can count on, though: You will usually get a downvote on Hacker News if you deviate from the mainstream. Previously you were in a discussion with me arguing with great confidence that the virus outbreak would be a distant memory in two months. I just looked it up; it was 61 days ago. https://news.ycombinator.com/item?id=22434499 https://news.ycombinator.com/item?id=22434499 You were using terms like "chicken littles" then. I'm all for reading different viewpoints, but if you're going to throw around insults like "doomer", you're not likely to get a civilised discussion going.
  - lurquer 7y ago
    Well, don’t known about your neck of the woods, but things are opening up here in two days. So, so I was a few days off in my estimate. Statistically insignificant deaths, no overwhelmed hospitals, no body bags in the hallways or rationing of ventilators. No millions of deaths now any other of the nonsense being predicted here two months ago. The talk now, as evidenced by the front page, is other topics... stimulus checks and the like. Are things as you feared two months ago?
- dboreham 7y ago
  With vs from is largely bogus.
nickledave 7y ago
Thank you, logged in to upvote this. These docs are committing basic errors. 6.6% infected has to be wrong. That just shows we're not testing enough to see the true, much lower, number. It's also biased because we're not randomly sampling--instead we're testing people that are already showing symptoms. Then when that inflated number goes in the denominator, it makes it look like the fatality rate is much lower.
- sixothree 7y ago
  How can this not be intentional?
  - James_Henry 7y ago
    It very well could have been. As someone pointed out in another comment, these guys talk very strongly about how someones "opinions about the data" might change when they aren't getting a paycheck. These two guys co-own several urgent care clinics are and thus probably suffering due to Covid-19 shelter-in-place orders.
  - kolanos 7y ago
    Which part? Taking what these doctors actually said out of context?
kolanos 7y ago
This is taken out of context. Those doctors were using aggregate testing and mortality statistics in New York and California and extrapolating. Sample sizes in the hundred of thousands. This Tweet is taking an anecdote about what they were seeing in their own practices out of context.
avsteele 7y ago
The LA county serology* showed 4.1%, with a confidence interval of 2.8% to 5.6%. That test that has a ~3 week lag so current #'s are even higher. Are you really prepared to say a claim of 6.6% is so misleading it needs to be removed? * http://publichealth.lacounty.gov/phcommon/public/media/mediapubhpdetail.cfm?prid=2328 http://publichealth.lacounty.gov/phcommon/public/media/media...
- DanBC 7y ago
  What's the selectivity and specificity of the test used in the study you quote?
  - avsteele 7y ago
    Selectivity is near 100% for tests where I've seen data.* I don't know about the specific one LA used but selectivity isn't the problem usually. The sensitivity is the issue, but that would tend to push the true infected % up not down. https://www.medrxiv.org/content/10.1101/2020.04.15.20066407v1 https://www.medrxiv.org/content/10.1101/2020.04.15.20066407v...
mixmastamyk 7y ago
What about the other countries, states, and cities they also compared?

by DanBC 7y ago

lurquer 7y ago

That was an obvious error in their reasoning. Not surprising, though. Idiotic (and, yes, such an error IS idiotic) fumbling with statistics is displayed on a daily basis. It doesn't matter what 'side' you are on: idiots abound. Doomer fallacies: Charts showing exponential growth of positives, without taking into account the exponential growth of tests... Flu-Bro fallacies: Antibody surveys that don't take into account false positives and make no mention of the expected number of positives in a population (both needed to make ANY sense of the results.) Doomer fallacies: No distinction between dying 'with' versus dying 'from'. Flu-Bro fallacies: Relying on death counts from the Chinese Communist Party. Doomer fallacies: Comparing 'positive' test rates between populations and countries with utterly different criteria for administering tests. etc. One thing you can count on, though: You will usually get a downvote on Hacker News if you deviate from the mainstream.

alexpetralia 7y ago
Yes I have historically found many doctors to not understand basic statistics, which has somewhat surprised me.
chimprich 7y ago
> One thing you can count on, though: You will usually get a downvote on Hacker News if you deviate from the mainstream. Previously you were in a discussion with me arguing with great confidence that the virus outbreak would be a distant memory in two months. I just looked it up; it was 61 days ago. https://news.ycombinator.com/item?id=22434499 https://news.ycombinator.com/item?id=22434499 You were using terms like "chicken littles" then. I'm all for reading different viewpoints, but if you're going to throw around insults like "doomer", you're not likely to get a civilised discussion going.
- lurquer 7y ago
  Well, don’t known about your neck of the woods, but things are opening up here in two days. So, so I was a few days off in my estimate. Statistically insignificant deaths, no overwhelmed hospitals, no body bags in the hallways or rationing of ventilators. No millions of deaths now any other of the nonsense being predicted here two months ago. The talk now, as evidenced by the front page, is other topics... stimulus checks and the like. Are things as you feared two months ago?
dboreham 7y ago
With vs from is largely bogus.

nickledave 7y ago

Thank you, logged in to upvote this. These docs are committing basic errors. 6.6% infected has to be wrong. That just shows we're not testing enough to see the true, much lower, number. It's also biased because we're not randomly sampling--instead we're testing people that are already showing symptoms. Then when that inflated number goes in the denominator, it makes it look like the fatality rate is much lower.

sixothree 7y ago
How can this not be intentional?
- James_Henry 7y ago
  It very well could have been. As someone pointed out in another comment, these guys talk very strongly about how someones "opinions about the data" might change when they aren't getting a paycheck. These two guys co-own several urgent care clinics are and thus probably suffering due to Covid-19 shelter-in-place orders.
- kolanos 7y ago
  Which part? Taking what these doctors actually said out of context?

kolanos 7y ago

This is taken out of context. Those doctors were using aggregate testing and mortality statistics in New York and California and extrapolating. Sample sizes in the hundred of thousands. This Tweet is taking an anecdote about what they were seeing in their own practices out of context.

avsteele 7y ago

The LA county serology* showed 4.1%, with a confidence interval of 2.8% to 5.6%. That test that has a ~3 week lag so current #'s are even higher. Are you really prepared to say a claim of 6.6% is so misleading it needs to be removed? * http://publichealth.lacounty.gov/phcommon/public/media/mediapubhpdetail.cfm?prid=2328 http://publichealth.lacounty.gov/phcommon/public/media/media...

DanBC 7y ago
What's the selectivity and specificity of the test used in the study you quote?
- avsteele 7y ago
  Selectivity is near 100% for tests where I've seen data.* I don't know about the specific one LA used but selectivity isn't the problem usually. The sensitivity is the issue, but that would tend to push the true infected % up not down. https://www.medrxiv.org/content/10.1101/2020.04.15.20066407v1 https://www.medrxiv.org/content/10.1101/2020.04.15.20066407v...

mixmastamyk 7y ago

What about the other countries, states, and cities they also compared?