- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
0
Posted on 4/17/20 at 9:36 pm to buckeye_vol
quote:
However, the first (and maybe only) FDA approved test by Cellex found that using that same methodology (LGM and LGG antibodies) estimates the sensitivity to be 93.8% and the specificity to be 96%.
If we just combine all of the studies of the negative tests (369/371 by Premier; 3/30 by the Stanford researchers; 240/250 by Cellex) then the specificity (639/651) was 98.16%, which means that 1.84% of the sample negative claws would be a false positive
Why are you assuming that their specificity is wrong?
And why are you only looking at false positives and not false negatives?
Posted on 4/17/20 at 9:36 pm to MidnightVibe
quote:South Korea and Germany were the two nations lauded for their test, trace, and containment and they’re both large enough (500,000+ tests) and far along enough in their progression to evaluate. And while their initial CFRs were well below, 1% initially which people used as evidence that the IFR was well below 1%, they are currently 3.1% and 2.2% and rising.
The fatality rate spread is nonsense.
So if their CFRs are that high, then even getting to a 1% IFR is a tough feat.
Posted on 4/17/20 at 9:38 pm to MidnightVibe
quote:Because they used unapproved Chinese tests that have since been banned because they lack a reliable study. Not to mention, the Chinese studies of their various tests specificity was closer to 90%, and the only FDA approved test and study was 96%. And that test was developed later.
Why are you assuming that their specificity is wrong?
And even more problematic than that, is that potential sampling bias of a self-selected sample based on Facebook recruitment.
This is a poor study, done by a group of scientists who had originally argued based on absurdly small and unrepresentative samples that the IFR was 0.01% (using NBA players and the cruise ships) which is now impossible since with over 37,000 deaths, that would mean there were more than 370 million infected, far more than our entire population.
This post was edited on 4/17/20 at 9:50 pm
Posted on 4/17/20 at 9:39 pm to RollTide1987
I have just been taking the reported numbers and multiply x100.
Bad news, most of us are going to get it.
Good news, it’s not as deadly as we thought.
Bad news, most of us are going to get it.
Good news, it’s not as deadly as we thought.
Posted on 4/17/20 at 9:44 pm to TutHillTiger
quote:Well then you have to do the same with the tests, and 3.575 million tests have been given. So multiplying that by 100, is 357.5 million tests, far more than the our population of about 331 million.
I have just been taking the reported numbers and multiply x100.
Posted on 4/17/20 at 9:52 pm to buckeye_vol
You said they aren’t accounting for false positives
Did you read the paper? It took my all of 5 minutes to find this
So you lied
Did you read the paper? It took my all of 5 minutes to find this
quote:
Third,we adjusted the prevalence for test sensitivity and specificity. Because SARS-CoV-2 lateral flowassaysare new,we applied three scenariosoftestkitsensitivity and specificity.Thefirstscenariousesthe manufacturer’s validation data (S1). The second scenario uses sensitivity and specificity from asample of37 known positive (RT-PCR-positive and IgGorIgMpositive on a locally-developed ELISA)and 30 known pre-COVID negativestestedonthekitatStanford(S2).Thethirdscenariocombinesthetwo collections of samples (manufacturer and local sample) as a single pooled sample (S3). Weuse the delta method to estimate standard errorsforthe population prevalence,which accountsforsampling errorand propagatesthe uncertainty in the sensitivity and specificity in each scenario.Amore detailed version ofthe formulaswe use in ourcalculationsisavailable in the Appendix to thispaper
So you lied
This post was edited on 4/17/20 at 9:53 pm
Posted on 4/17/20 at 9:55 pm to WaWaWeeWa
quote:I was wrong. It was done by Harvard scientists (Marc Lipsitch). He tweeted it out when these Stanford scientists wrote that ridiculous OPED estimating father IFR to be 0.01%, but now I can’t find it.
When was the Yale study? I know Fauci said that before antibody testing.
Posted on 4/17/20 at 9:58 pm to WaWaWeeWa
quote:Yes. I even used the data (369/371 from manufacturer; 30/30 in their control test) that they presented in their study in my estimation:
You said they aren’t accounting for false positives
Did you read the paper? It took my all of 5 minutes to find this
quote:If I included an even less favorable study of the Chinese antibody tests (and this was a Chinese test) that had a specificity around 90%, it would have been been worse.
If we just combine all of the studies of the negative tests (369/371 by Premier; 3/30 by the Stanford researchers; 240/250 by Cellex) then the specificity (639/651) was 98.16%, which means that 1.84% of the sample negative claws would be a false positive.
This post was edited on 4/17/20 at 9:59 pm
Posted on 4/17/20 at 10:01 pm to buckeye_vol
Yes but you are only accounting for false positives, why aren’t you talking about false negatives?
They said the sensitivity was around 80%
They said the sensitivity was around 80%
Posted on 4/17/20 at 10:09 pm to WaWaWeeWa
quote:Because with such a a low infection rate, false positives will dwarf false negatives, even if the specificity is 98.5% and the sensitivity is only 80%.
Yes but you are only accounting for false positives, why aren’t you talking about false negatives?
They said the sensitivity was around 80%
For example, with a 98.5% specificity and an 80% sensitivity, and with 0.5% infection rate then there will be 0.1% false negatives and 1.49% false positives.
It’s called the False Positive Paradox for this very reason.
This post was edited on 4/17/20 at 10:10 pm
Posted on 4/17/20 at 10:18 pm to buckeye_vol
Well we will see soon enough after peer review. It should be fairly easy to determine if their statistical analysis is accurate.
Posted on 4/17/20 at 10:31 pm to buckeye_vol
quote:
Well then you have to do the same with the tests,
Huh?
Posted on 4/17/20 at 10:34 pm to buckeye_vol
quote:
And while their initial CFRs were well below, 1% initially which people used as evidence that the IFR was well below 1%, they are currently 3.1% and 2.2% and rising.
Really?
LINK
Posted on 4/17/20 at 10:43 pm to WaWaWeeWa
quote:The problem is that unless the peer-reviewers are aware that the manufacturer imported an unapproved and now banned test, and are unaware of the other studies that indicate far worse specificity of approved tests, then they have to take the manufacturer at their word, unless they request the researchers to do a more thorough control analysis than their 30 sample analysis.
Well we will see soon enough after peer review. It should be fairly easy to determine if their statistical analysis is accurate.
But the peer-review process usually assumes that the manufacturer’s data is valid, so I doubt they’ll question it.
But the biggest problem, that’s impossible to account for (as they even note) is the extent that they’re sample is biased (self-selected and recruited from Facebook). I suspect that those who agreed to do the antibody test were more likely to be from those who had a higher probability of being positive (e.g., confirmed positive; displayed symptoms but couldn’t get tested; etc.).
And I think that they’re adjustments to reflect the population (e.g., sample was 63% female, but population is only 50% female), which increased the preference from 1.5% to 2.81% (an 87% increase) actually exacerbated this problem.
Of course, last week when I learned that they were doing this study, and before I even realized it came from those arguing that the IFR was an impossible 0.01%, I said that these studies would reveal a high prevalence that greatly overestimated the prevalence. And I was assuming a random sample at that.
Until we can get tests that have gone through a thorough, independent analysis indicating specificity well above 99% or at least until the spread of the illness is large enough that the base rate is not so low, these studies are pretty worthless at estimating the true prevalence.
At the very least, they should assume the accuracy is at the lower bound, especially in a field study (more potential errors than a thorough lab study for that purpose).
Posted on 4/17/20 at 10:50 pm to MidnightVibe
quote:Again. They tested a sample from a small town that had a huge outbreak, and like this Stanford study, there is a potential sampling bias in those that self-selected to participate. Plus they also assume the specificity of these tests, likely from labs that haven’t gone through any independent analysis, is far greater than it likely is.
Really?
But since they generalized (which isn’t valid anyways) it to the CFR at the time, which was below 2% and is now over 3%, a 55% increase, their generalization of the IFR has now increased by 55% as well. It’s now above 0.6% and rising.
And it will likely continue to rise since deaths lag cases so the CFR rises (from just above 1% at the end of March to just above 3% now).
This post was edited on 4/17/20 at 10:57 pm
Posted on 4/17/20 at 11:07 pm to buckeye_vol
quote:
But the biggest problem, that’s impossible to account for (as they even note) is the extent that they’re sample is biased (self-selected and recruited from Facebook). I suspect that those who agreed to do the antibody test were more likely to be from those who had a higher probability of being positive (e.g., confirmed positive; displayed symptoms but couldn’t get tested; etc.).
I actually doubt it. This is a serological test. Self-selection would be quite difficult.
Posted on 4/17/20 at 11:12 pm to buckeye_vol
quote:
Again. They tested a sample from a small town that had a huge outbreak, and like this Stanford study, there is a potential sampling bias in those that self-selected to participate. Plus they also assume the specificity of these tests, likely from labs that haven’t gone through any independent analysis, is far greater than it likely is.
But since they generalized (which isn’t valid anyways) it to the CFR at the time, which was below 2% and is now over 3%, a 55% increase, their generalization of the IFR has now increased by 55% as well. It’s now above 0.6% and rising.
And it will likely continue to rise since deaths lag cases so the CFR rises (from just above 1% at the end of March to just above 3% now).
USS Roosevelt. Entire ship tested. 660 positive for covid. 60% asymptomatic. 1 death.
You're welcome.
Posted on 4/17/20 at 11:14 pm to buckeye_vol
quote:
But the peer-review process usually assumes that the manufacturer’s data is valid, so I doubt they’ll question it.
Yeah, I'm sure they aren't as smart as you.
Posted on 4/17/20 at 11:16 pm to MidnightVibe
Page 9 of 10
Popular
Back to top
Follow TigerDroppings for LSU Football News