Asking a girl on a second date.

less arguing more explaining what was wrong with my math

if you test 300 million known nonmormon ppl with 95% accuracy, it doesnt matter if you get "5% false negatives" obviously so you only look at the false positives.

same argument for looking at the mormon thing.

yes, the question lacked a little definition but once clearly defined (as i did in my little math model which essentially matches what raven showed) i dont see whats wrong there.

if there is something wrong, then id love for someone to explain it WITH math
<3 thx

ALSO MORMONS ARE CRAZY :lolwut:
 
The correct way to phrase that would be "major in math at a liberal arts college," assuming he goes to Dartmouth.
 
Eggi said:
if you test 300 million known nonmormon ppl with 95% accuracy, it doesnt matter if you get "5% false negatives" obviously so you only look at the false positives.
Click to expand...
95% accuracy means 5% of the time the test is wrong, ie gives false answers. it gives you no information where those false answers are, just the total. the answer to the question given depends on false positives.

Total False = FP + FN = 5% * population

a) this means you can't know FP without knowing more information.
b) FP and FN are interrelated by the equation above, so you cannot just look at FP.

depending on the ratio of FP/FN, the answer to the question could range from 0% to 100%.
 
Yeah, he was right too, but I was looking for a second opinion to be sure.

Thanks anyway. (by the way, I totally FUBARed the problem in my original post)
 
SegaRob said:
Well when someone does that you have three options:
1) Make up a wrong answer.
2) State what information you would need to solve the problem.
3) Look up the information from 2) and actually solve the problem.

I took option 2. Just because he left out critical information doesn't justify people taking option 1.
Click to expand...

yeah in an interview, not a tw brain teaser :lol:

opsayo said:
ayz,pretty much all higher level math classes are all proofs

was minoring at your college pretty easy?
Click to expand...

obviously you did proofs in calc classes, although not in any diffeq classes i took

i'm not talking about basic proofs. my roommate was a math major and took that class, they did a bunch of obscure abstract proofs like "prove numbers exist" or some stupid shit, the kid would do his homework and have 5-6 pages for a single proof.
 
SegaRob said:
95% accuracy means 5% of the time the test is wrong, ie gives false answers. it gives you no information where those false answers are, just the total. the answer to the question given depends on false positives.

Total False = FP + FN = 5% * population

a) this means you can't know FP without knowing more information.
b) FP and FN are interrelated by the equation above, so you cannot just look at FP.

depending on the ratio of FP/FN, the answer to the question could range from 0% to 100%.
Click to expand...

so, that really makes no sense - you cant determine when something messes up and still gets the right answer, unless its explicitly stated. Every problem i've ever tackled with accuracies doesnt sit there and talk about well its 95% accurate, but then 2% of that inaccuracy still yields the right answer...cuz you know, then itd be 97% accurate

:shrug:

maybe its cuz im an engineer and i only deal with real world problems :)
 
Eggi said:
so, that really makes no sense - you cant determine when something messes up and still gets the right answer, unless its explicitly stated. Every problem i've ever tackled with accuracies doesnt sit there and talk about well its 95% accurate, but then 2% of that inaccuracy still yields the right answer...cuz you know, then itd be 97% accurate

:shrug:

maybe its cuz im an engineer and i only deal with real world problems :)
Click to expand...
well you're obviously not a very good one if you can't understand this simple mathematical concept. and this is real world problem - go look up Receiver Operating Characteristic (ROC).

there are TWO types of right answers and TWO types of wrong answers. you can make a 2x2 table of the possible outcomes:

Code: 
              |         Mormon      |       Not Mormon
------------------------------------------------------
Test Positive |    [b]True Positive[/b]    |    [b]False Positive[/b]
------------------------------------------------------
Test Negative |   [b]False Negative[/b]    |    [b]True Negative[/b]
We can abbreviate the outcomes as TP, FP, FN, TN.

We were given accuracy (acc) as 0.95, which means the true results are 95% of the total outcomes:

acc = (TP + TN) / (TP + TN + FN + FP) = 0.95

Where (TP + TN + FN + FP) is the entire sample space, or population.

Hence,

TP + TN = 0.95*population

AND:

FP + FN = 0.05 * population (which I stated before)

These FP and FN are both WRONG answers, but depending on the relative number of FP to FN, the answer to raven's question is different. The parameter raven asked for is known as the true positive rate, TPR, which is defined as the number of true positives out of the total number of positives identified by the test.

Mathematically that can be written as:

TPR = FP / (TP + FP)

However, FP is not uniquely defined for a given value of acc. A second parameter that is independent of acc must be defined in order to determine FP and hence TPR. Based on the value of FP, TPR can range from 0 to 1, the full domain for TPR.

And I wasn't even a math major.
 
You must log in or register to reply here.
Back
Top