Perhaps the BEST B1G Forum anywhere, here at College Football Fan Site, CFB51!!!

And this is the problem with this whole covid thing reliable individuals (CD & Bwarb) are often in conflict/disagreement.Wanna believe somebody but Wuhan, I mean who?

And this is the problem with this whole covid thing reliable individuals (CD & Bwarb) are often in conflict/disagreement.Wanna believe somebody but Wuhan, I mean who?

I wasn't disagreeing with CD. I was clarifying.

The tests aren't reliable for predicting whether YOU have antibodies if YOU get a positive result. But that's not a problem with the test, it's a problem with having such a low prevalence of antibodies in the population that you can't be sure whether your test was a real positive or a false positive. 

Think of it this way: 

Prevalence: 1% of the population has antibodies
Test A: 5% false positive rate: You will get about 6% actual positive result from the test. Which means for any individual person who tests positive, they have only a 17% likelihood that they actually have antibodies.
Test B: 1% false positive rate: You will get about a 2% actual positive results from the test. Which means for any individual person who tests positive, they only have an 50% chance that they actually have antibodies.

Prevalence: 5% of the population has antibodies

Test A: 5% false positive rate: You will get about 10% actual positive result from the test. Which means for any individual person who tests positive, they have only a 50% likelihood that they actually have antibodies.
Test B: 1% false positive rate: You will get about a 6% actual positive results from the test. Which means for any individual person who tests positive, they only have an 83% chance that they actually have antibodies.

Prevalence: 50% of the population has antibodies

Test A: 5% false positive rate: You will get about 55% actual positive result from the test. Which means for any individual person who tests positive, they have only a 91% likelihood that they actually have antibodies.
Test B: 1% false positive rate: You will get about a 51% actual positive results from the test. Which means for any individual person who tests positive, they only have an 98% chance that they actually have antibodies.

So the key is that even for a very accurate test, and I think most antibody tests are around the specificity of Test B in the above scenarios, the actual predictive power of a GOOD test depends highly on the prevalence of the condition you're testing for, and the prevalence is too low right now to give as much predictive power as we would like. 

(Note above: those numbers aren't perfect, because I didn't take into account the fact that a false positive rate only affects the non-positive folks (i.e. 5% false positive when you have 50% prevalence should only give you 52.5% actual positive rate, because you can only get false positives on the folks who aren't true positives (5% * 50% = 2.5%). However I left those out because I didn't want to mess with the false negative rate, which also has to be taken into account.)

I wasn't disagreeing with CD. I was clarifying.

The tests aren't reliable for predicting whether YOU have antibodies if YOU get a positive result. But that's not a problem with the test, it's a problem with having such a low prevalence of antibodies in the population that you can't be sure whether your test was a real positive or a false positive.

Think of it this way:

Prevalence: 1% of the population has antibodies
Test A: 5% false positive rate: You will get about 6% actual positive result from the test. Which means for any individual person who tests positive, they have only a 17% likelihood that they actually have antibodies.
Test B: 1% false positive rate: You will get about a 2% actual positive results from the test. Which means for any individual person who tests positive, they only have an 50% chance that they actually have antibodies.

Prevalence: 5% of the population has antibodies

Test A: 5% false positive rate: You will get about 10% actual positive result from the test. Which means for any individual person who tests positive, they have only a 50% likelihood that they actually have antibodies.
Test B: 1% false positive rate: You will get about a 6% actual positive results from the test. Which means for any individual person who tests positive, they only have an 83% chance that they actually have antibodies.

Prevalence: 50% of the population has antibodies

Test A: 5% false positive rate: You will get about 55% actual positive result from the test. Which means for any individual person who tests positive, they have only a 91% likelihood that they actually have antibodies.
Test B: 1% false positive rate: You will get about a 51% actual positive results from the test. Which means for any individual person who tests positive, they only have an 98% chance that they actually have antibodies.

So the key is that even for a very accurate test, and I think most antibody tests are around the specificity of Test B in the above scenarios, the actual predictive power of a GOOD test depends highly on the prevalence of the condition you're testing for, and the prevalence is too low right now to give as much predictive power as we would like.

(Note above: those numbers aren't perfect, because I didn't take into account the fact that a false positive rate only affects the non-positive folks (i.e. 5% false positive when you have 50% prevalence should only give you 52.5% actual positive rate, because you can only get false positives on the folks who aren't true positives (5% * 50% = 2.5%). However I left those out because I didn't want to mess with the false negative rate, which also has to be taken into account.)

Everyone should get tested, and everyone should do it multiple times.

and what will all this testing accomplish?

back in March when there wasn't nearly enough tests to meet demand, I was told if we had enough tests we would win this battle.  Testing was the key to flattening the curve.

Well, it seems there are ample tests and the curve seems to be on an uptick most places

I don't know what this means.  When I get tested, it's either accurate or it's not. 

I described it above. All tests have error. 

Let's assume a test has a 1% false positive rate and a 5% false negative rate, and we go to the ultimate extremes.

If a population has 0% infection, i.e. not a single case, and you test that entire population, you'll get a 1% positive rate. Obviously for those 1%, the test is inaccurate, because we already stipulated 0% infection.

If a population has a 100% infection rate, and you test that entire population, you'll get only a 95% positive rate. Obviously for the 5% that tested negative, the test is inaccurate, because we already stipulated 100% infection.

Those are the easy cases...

Now for the hard one. 

If a population has a 50% infection rate, you know that you'll get an extra 0.5% test result from the 1% false positive rate on the half of the people who aren't infected, and you'll lose 2.5% of your true positives for the false negative rate on the half of the people who ARE infected, so you should expect a 48% result when testing the entire population. 

If your population is 100,000 people, that means that you have 500 people for whom the positive result was inaccurate, and 2500 people for whom the negative result was inaccurate. 

You know 50,000 people have the infection and 50,000 don't, but 3,000 people got an inaccurate test result. And none of the people can say with 100% certainty whether they're in the group which had an accurate result or an inaccurate result. 

This is why we distinguish between the accuracy of a test, and its predictive power. A very accurate test can have low predictive power depending on prevalence of infection.

And it's MUCH more important in this case for false positives, because you don't want those 500 people to think they're invincible and can't get re-infected, because they don't realize they were never infected in the first place and have no immunity. 

Everyone should get tested, and everyone should do it multiple times.

I do think we'd learn a lot if we had much more antibody testing, and if it was repeated multiple times to weed out false positive/negative results and to track the change over time towards herd immunity.

But that's a blood test. I'm not volunteering for anyone to jab a swab into my brain for the PCR testing unless I have legitimate reason to think I've been exposed. 

and what will all this testing accomplish?

back in March when there wasn't nearly enough tests to meet demand, I was told if we had enough tests we would win this battle.  Testing was the key to flattening the curve.

Well, it seems there are ample tests and the curve seems to be on an uptick most places

When this started, we didn't have enough tests to find the infected individuals and perform contact tracing to test anyone we knew they came in contact with. That is a very important aspect of slowing the spread. That's how places like South Korea were able to manage it. 

Unfortunately I believe our infection rate is too high at this point for contact tracing to do all that much good, particularly as we've reopened. However as utee's pointed out, when you at least have a known point of potential spread (his hair stylist) who has a client list of known patrons, you can contact trace. Right now we're trying to figure out if we're in the same boat--one of the doctor's at my wife's practice might have exposure due to a doctor she shares an office with at her hospital position who tested positive, and is in quarantine. If she likewise shows symptoms or tests positive, my wife will need to get tested, and then likely so would I, even though we don't think my wife came into close enough contact with the doctor to get it. 


That said, I don't think even testing / contact tracing will completely stop the spread. South Korea has been seeing an uptick in new cases ever since they started reopening. Hopefully they'll at least be able to keep it minimal and not allow it to explode.