[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 1.39 MB, 2000x1338, 1371799747490.jpg [View same] [iqdb] [saucenao] [google]
5849659 No.5849659 [Reply] [Original]

in case /b/ will fail me:

quick question, need to know this for my gf

a pregnancy test has 95% accuarcy
if she takes two test, how high is the accuarcy now?
(i am lucky, because both tests negative)

pic unrelated

>> No.5849664

of course, /b/ failed me:

>>>/b/488362887

>> No.5849666

95%
taking more tests won't prove anything

>> No.5849668

It's just as accurate than if you only took one test.

This is fucking textbook shit.

>> No.5849667

>>5849659
95%

>> No.5849673

>>5849659
0.95[accuracy] ^ (2 [tests]) = 0.9025 [accuracy]

surprising results, if I do say so myself.

it turns out performing an additional test will actually increase her chances of being pregnate by almost 5%.

>> No.5849672

do you even know what 95% accuracy means?
you'll have to be more precise than "95% accuracy", we need to know what is the probability the test will fail even if the person is pregnant, and the probability the test will be positive if the person is not pregnant.
I'll develop on the reason why "95%" isn't sufficient

>> No.5849674

You can't tell with certainty. The 5% failure can happen for multiple reasons that screw up any meaningful numbers. A certain amount of failure occurs through a defective product. Another type of failure is from the woman's chemistry. Taking two tests will help eliminate the errors of a defective product, but it's still the same woman with the same chemistry "error". This "error" will salt the data.

>> No.5849678

what about this

>>>/b/488363845

>> No.5849680

>>5849678

sounds logical to me

>> No.5849682

>>5849678

Yeah that's correct. The probability of failure for muiltiple tests is (1-P)^n.

So the probability of success is 1 minus that, or 1-(1-.95)^2 or 1/400.

>> No.5849683

>>5849682

Whoops, I mean the probability of failure is 1/400, the probability of success (accuracy) is 399/400 or 99.75%

>> No.5849684

It depends on what "95% accurate" means. Does it mean 5% chance of false positive? 5% chance of false negative? Both? Does it mean it's 100% accurate in 95% of circumstances? 95% accurate in 100% of circumstances?

Without knowing those things, you can't say for sure.

>> No.5849692

It is either the 99.75 or 95%. The 99.75 would come from a defective product. This would be like if there were popsicle sticks in pregnancy tests 5% of the time. If you randomly grab two tests the chance of them both being popsicle sticks is 1/400. However, it is 95% if the failure rate isn't random and is based on something else like your woman's actual body chemistry

>> No.5849697

>>5849684

Just call it "success" and "failure".

So if the girl is pregnant and it tests positive that is a "success", if it is negative it's a "failure".
Similarly if she is not pregnant and it tests positive that is a "failure" and a negative test is a "success"

All failures have a 5% chance, all successes have a 95% chance, assuming we know beforehand whether the girl is pregnant.

>> No.5849706

>>5849697

That's very unlikely though that the chance of false positive is exactly the same as the chance of false negative.

>> No.5849729

>>5849672
imagine that if your gf is pregnant, the test is positive 95% of the time, and that if she is not, the test is negative 95% of the time.
Also imagine you impregnated her with a probability p.
then the probability P your gf is pregnant knowing the test is negative is:
P= 0.05*p/(0.05*p+ 0.95*(1-p)) = 0.05p/(0.95-0.9p)
if p = 10%, the test will be negative even if she is pregnant with a probability of 6%.

but if p=80%, the test will be negative more than 17% of the time.

if p=90%, the test will be negative more than 32% of the time.
and it keeps growing.


that's one of the applications of Baye's theorem.

>> No.5849926
File: 6 KB, 104x100, 1357172518807.jpg [View same] [iqdb] [saucenao] [google]
5849926

>>5849673

this

>> No.5849963
File: 57 KB, 899x489, jp_166.jpg [View same] [iqdb] [saucenao] [google]
5849963

>95% accurate result rate
>5% false result rate

.05*.05 = .0025

>0.25% false result rate after two tests

>99.75% accurate result rate after two tests

>> No.5849986

>>5849659
>in case /b/ will fail me:

>/b/ will fail

Fixed

>> No.5849994

lol at the babies who do not know probability

the probability is .95 + (.05*.95)

=0.9975

95% isnt the fallability of just the test though, so taking two back to back is not a good idea. Also false positives are very rare

>> No.5850002

>>5849692
This answer is as accurate and complete as it gets. The thread should have stopped there. OP's gf has between 0.25% and 5% chances of being pregnant.

>> No.5850007

Can't give a definitive answer because some people could be more predisposed to false results than others. In fact, I'd be highly surprised if this wasn't the case.

>> No.5850071

Newsflash: If you test positive, it's quite reliable, if you test negative not so much.

>> No.5850078

>>5849659
pregnancy tests dont follow simple statistics there are various factors that could mask the test result if your gf thinks she is pregnant she should go see a gynacologist.