
Epidemiology probability question.
Suppose a test for a disease is 99% accurate for positive results. This means there is a 1% false positive rate.
Suppose 1% of the entire population has the disease.
You get tested. The test is positive.
What is the probability that you have the disease?
.
Stout beers under trees, please.

Originally Posted by QuiQuaeQuod
Suppose a test for a disease is 99% accurate for positive results. This means there is a 1% false positive rate.
Suppose 1% of the entire population has the disease.
You get tested. The test is positive.
What is the probability that you have the disease?
.99. The number of people that have it is not relevant

Originally Posted by QuiQuaeQuod
Suppose 1% of the entire population has the disease.
What is the probability that you have the disease?
Preposterous. Your supposition is flawed.
Dr. Trump has ensured us the virus will magically disappear. Thus any percent of 0% = Ø

Originally Posted by DaveG
.99. The number of people that have it is not relevant
Wrong, and wrong.
Most people get it wrong, btw. And 99 is a very common answer. So are 98, and 1. Which are also wrong.
.
Stout beers under trees, please.

Originally Posted by QuiQuaeQuod
Wrong, and wrong.
Most people get it wrong, btw. And 99 is a very common answer. So are 98, and 1. Which are also wrong.
can you show the math?

Originally Posted by DaveG
can you show the math?
Yes.
And eventually I will.
.
Stout beers under trees, please.

.99 just got hit by a bus walking out of the clinic. Time to recrunch the numbers.
Where's my abacus?

Originally Posted by QuiQuaeQuod
Yes.
And eventually I will.
Bayes's Theorem

Originally Posted by DaveG
Bayes's Theorem
I won't need to refer to that to show my math, but yes that will get you there.
.
Stout beers under trees, please.

50%, either you do, or you don't, same way I figure all my odds.
Disclaimer: I am NOT good in math
Of course I'm sure...that doesn't mean I'm right.....
"There's no sense being stupid unless you show it."

Originally Posted by QuiQuaeQuod
Wrong, and wrong.
Most people get it wrong, btw. And 99 is a very common answer. So are 98, and 1. Which are also wrong.
The only premise I can think of that would make his answers wrong is if "false positive" in not a confirmation of 'negative'.
Is that why you say he's wrong?
My understanding of the term is that it indeed is the same as negative.

Originally Posted by Jay Strongbow
The only premise I can think of that would make his answers wrong is if "false positive" in not a confirmation of 'negative'.
Is that why you say he's wrong?
My understanding of the term is that it indeed is the same as negative.
False positive = test says you have it, but you really don't.
True negative = test says you don't have it, and you don't.
Same disease status, different test results.
.
Stout beers under trees, please.


Originally Posted by QuiQuaeQuod
Suppose a test for a disease is 99% accurate for positive results. This means there is a 1% false positive rate.
Suppose 1% of the entire population has the disease.
You get tested. The test is positive.
What is the probability that you have the disease?
Well, I'd think the probability of being infected is 50% (two possible outcomes, infected: Y/N) ...and the probability of your individual test result being correct is 25% (a further 2 outcomes, Test correct: Y/N).
EDIT: And I'd be wrong...Interesting topic.
Last edited by Marc; 3 Weeks Ago at 04:38 PM.
"Refreshingly Unconcerned With The Vulgar Exigencies Of Veracity "

Originally Posted by Marc
Interesting topic.
I agree, of course. For many reasons. But mainly because what seems so obvious.... isn't.
.
Stout beers under trees, please.

Originally Posted by QuiQuaeQuod
What is the probability that you have the disease?
The math (barely math, so you'll have no problem) is below, Highlight or quote to see it.
The math is easy to do percentages with if we think about 100 tests. Do 100, and you will find 1 case, a true positive. That's how many people have it, 1%. You will also have 1 false positive. So 100 tests, 2 will be positive.
So your personal positive test? 5050. Half of all positives will be false, half will be true. Under the specified conditions.
.
Stout beers under trees, please.

I'm going with my answer.

Originally Posted by QuiQuaeQuod
The math (barely math, so you'll have no problem) is below, Highlight or quote to see it.
The math is easy to do percentages with if we think about 100 tests. Do 100, and you will find 1 case, a true positive. That's how many people have it, 1%. You will also have 1 false positive. So 100 tests, 2 will be positive.
So your personal positive test? 5050. Half of all positives will be false, half will be true. Under the specified conditions.
Huh, not where I thought that was going:
https://brownmath.com/stat/falsepos.htm
"Refreshingly Unconcerned With The Vulgar Exigencies Of Veracity "

So .... pretty much the tests are useless.

Originally Posted by duriel
So .... pretty much the tests are useless.
Not really,, A test is just way less predictive than people think for some things that are relatively rare. A single positive test does increase your probability from 1% up to X%, and that's a big jump.
But you will still want a retest to really confirm. Testretest is very useful.
.
Stout beers under trees, please.

For all new tests, which have not went through a series of rigorous testing and evaluation (which is what we have), actually 50% of the population support that it's nothing more than the flu.
A test > useless by its self.
2 tests > maybe a little better.
more tests > could be a little better, but basically useless.
no pulse> pretty good test.
This is what we have for Cov19 tests.

Originally Posted by duriel
For all new tests,
...
A test > useless by its self.
2 tests > maybe a little better.
I'm not talking about COVID, I am talking in general. But the risk of a false positive is a lot lower in NYC, given a much higher than 1% population prevalence. That's how the math works. The more that is out there, the less likely a given positive test is to be a false positive.
A single positive test moves your chance of having it up by 49%. I think that is useful information for an individual to do something about. Like keep away from others until a retest, be aware of even "minor" symptoms of the disease, etc.
The chance of TWO false positives for the same person at the 1% (1/100) false positive rate is... 1/10,000. So double test 10,000 people, you get 101 double positives, one of which is false. 1/101 = very slightly over 99% sure any given double positive is a true positive.
Not really a LITTLE better, imo.
.
Stout beers under trees, please.

Originally Posted by QuiQuaeQuod
The math (barely math, so you'll have no problem) is below, Highlight or quote to see it.
In my 'math' the 1 true positive person was given the 1 false positive test. Thus the result was 1 positivefalsepositive.
What's the probability of that?

Originally Posted by tlg
In my 'math' the 1 true positive person was given the 1 false positive test. Thus the result was 1 positivefalsepositive.
What's the probability of that?
What volume is your sky set on?
IOW, that makes no sense. There is no such thing, by definition, as a person who is positive getting a false positive test result. False positives require a negative status to exist.
.
Stout beers under trees, please.

Originally Posted by QuiQuaeQuod
I'm not talking about COVID,
I am talking about cov19.
And it you read the paper marc posted, without a thorough evaluation, many people give false results due to too many things not associated with the test. Like maybe if you had coffee that morning.
Similar Threads

By PaulCL in forum General Cycling Discussion
Replies: 2
Last Post: 08232006, 02:33 PM

Replies: 4
Last Post: 07222005, 02:49 PM

By KenL in forum Components, Wrenching
Replies: 1
Last Post: 07052005, 05:11 AM

By Tri_Rich in forum Politics Only
Replies: 1
Last Post: 02212005, 10:27 AM

By andy02 in forum General Cycling Discussion
Replies: 1
Last Post: 03192004, 07:55 PM
Posting Permissions
 You may not post new threads
 You may not post replies
 You may not post attachments
 You may not edit your posts

Forum Rules

