Bayesian probability on disease

170 Views Asked by At

There is a rare disease that only happens to 1 out of 100,000 people. A test shows positive 99% of times when applied to an ill patient and, 1% of times when applied to a healthy patient. Please answer the following questions.

  • What is the probability for you to have the disease given that your test result is positive?

  • What is the probability for you to have the disease when you did two tests and both of them show positive? Assume that two tests are conducted independent.

  • Assume that the patient keeps on trying the test, what is the minimum number of tests that the patient has to try to be 99% percent sure that he is actually ill? Assume that all tests are conducted independently.

$$ \Pr(D\mid +)= { \Pr(+\mid D)\Pr(D) \over (\Pr(+\mid D)\Pr(D)+\Pr(+\mid \bar D)\Pr(\bar D)) } $$ On substituting we get approximately $0.0989 %$ But for the second question if he/she tested positive for both times The probability should increase of him having the disease But when we multiply $0.989\times 0.989=0.978$ is the probability he has the disease

Please help me where I am going wrong

1

There are 1 best solutions below

2
On

Your answer for $P(D|{+})$ in the comments of about 0.1% is correct. But surely you know that multiplying two numbers less than 1 will give you another number which is less than either of them, so your method cannot be right.

We want $P(D|{+}{+})$, the probability of $D$ given two positive test results. You can use Bayes' theorem again to get $$ P(D|{+}{+})={P({+}{+}|D)P(D)\over P({+}{+})} $$ But we need to think about what the values of $P({+}{+}|D)$ and $P({+}{+})$ are. We're told that the results are independent, so $P({+}{+}|D)=P({+}|D)P({+}|D)$.

The value of $P({++})$ is given by $$P({+}{+})= P({+}\mid D)^2P(D)+P({+}\mid\bar D)^2P(\bar D) $$ since the tests are independent of each other, but conducted on the same person with the same chances of having or not having the disease. So the answer is

$$ P(D|{+}{+})={P({+}|D)^2P(D)\over P(++)} $$ You already have $P({+}|D)$ and $P(D)$ from your answer, and you can calculate $P({+}{+})$ as shown above, so just plug the values in. It comes out to about 9%.