Discrete math:
A company uses computer vision software for quality inspection of its products. Over all products, 1 out of 10 is defective. The software labels non-defective products to be correct in 17 out of 18 cases. Defective products are (falsely) labeled to be correct in 1 out of 5 cases. Given that a product is labeled to be correct, what is the probability that it is actually defective?
5/64 |
||
2/87 |
||
0.0018 |
||
0.078 |
The producer of a laboratory test for detecting a virus estimated the following table of conditional probabilities for the test. The test can be either true (T) or false (F) and a patient may either carry the virus (+) or not carry the virus (-). In addition, we know that the prevalence for a virus infection (probability to have the virus given no prior knowledge) is p(+) = 0.002.
p( test | patient ) | patient + | patient - |
test T | 0.93 | 0.81 |
test F | 0.07 | 0.19 |
1.00 | 1.00 |
What is the probability to carry the virus if the test is positive?
0.0389 |
||
0.0313 |
||
0.0427 |
||
0.0023 |
1)
probability that it is actually defective given labeled correct
=P(defective and labeled correct)/P( labeled correct )
=(1/10)*(1/5)/((1/10)*(1/5)+(9/10)*(17/18))
=2/87
2)
probability to carry the virus if the test is positive =(0.002*0.93)/(0.002*0.93+(1-0.002)*0.81)=0.0023
Discrete math: A company uses computer vision software for quality inspection of its products. Over all...