From Hilary Mason's machine learning course at O'Reilly.
10000 people in all
100 people have disease (+D)
9900 people do not have disease (-D)
A test for the disease is 99% effective;
i.e., of the 100 people who have disease, 99 of them test positive (+T) and 1
tests negative (-T).
and of the 9900 people who don't have the disease, 99 (1%) test positive(+T) and
9801 (99%) test negative (-T).
Question: given than a person tests positive, what is the probability that
person has the disease?
+D -D
+T 99 99
-T 1 9801
Answer (from first rown) 50% (99 are in the (+T and +D) group), and
(99+99) are in the +T group. so 99/(99+99) = 50% is the answer.
From Hilary Mason's machine learning course at O'Reilly.