Confused on how to apply Bayes Th. to this problem.

You have a machine that can identify a disease in 75% of cases, and in 95% of cases the machine is able to correctly indicate that a person does NOT have a disease. What is the probability of a machine incorrectly saying a person has a disease when they do not, and what is the probability of the machine incorrectly saying a person does not have a disease when they actually do?


Assume M stands for the machines predictability. Assume D stands for a person having the disease.

P(M)=0.75 which means the machine correctly predicts 75% of the time, P(M')=0.25 which means the machine incorrectly predicts 25% of the time.

P(M|D')=0.95 meaning a person does not have a disease and the prediction is positive.

I need to first find P(M'|D) is the probability of a person having the disease but the prediction is wrong. I then need to find P(M'|D') which is person not having disease and prediction is wrong.

I am stuck here. Is my thought process correct? Do I not need to have information of P(D) to proceed?

  • 1
    $\begingroup$ I don't see the need for Bayes' Theorem or conditional probabilities. Why aren't the answers by definition $100\%-95\%=5\%$ for the first question and $100\%-75\%=25\%$ for the second? Am I misunderstanding what you mean by the first sentence of the second paragraph? If so, could you please clarify what you do mean? $\endgroup$
    – joriki
    Sep 10 '18 at 19:18
  • $\begingroup$ Related: math.stackexchange.com/questions/32933/… $\endgroup$
    – Henry
    Oct 8 '20 at 10:13

I think you are going astray because your interpretation of the problem statement is a bit off. When the problem states:

You have a machine that can identify a disease in 75% of cases

This is conditional on the disease being present, i.e., $P(M|D) = 0.75; P(M'|D) = 0.25.$

Then, you won't need need $P(D)$ to answer the two questions that were posed - they can be determined by thinking through what $P(M'|D')$ and $P(M'|D)$ represent.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.