Week6-2Bayesian theorem

Bayes’ theorem

• Formula for joint probability
  
  • $P(A,B) = P( B \mid A)P(A) = P(A \mid B)P(B)$
• Therefore

$$P(A \mid B) = \frac{P(B \mid A)P(A)}{P(B)}$$

Example

We have the following table.

What is $p( disease \mid positive)$?

$$P( disease \mid positive) = \frac{P( positive \mid disease) P(disease)}{P(positive)}\\ P( \neg disease \mid positive) = \frac{P( positive \mid \neg disease) P(\neg disease)}{P(positive)}$$

We don’t care what value $P(positive)$ is, as long as it is larger than 0.

$$\frac{P(disease \mid positive)}{P(\neg disease \mid positive)}= \frac{P( positive \mid disease) P(disease)}{P( positive \mid \neg disease) P(\neg disease)}$$

So $P(disease)$ is crucial for the final ratio. Suppose $P(disease) = 0.001$, and we would have

$$\frac{P(disease \mid positive)}{P(\neg disease \mid positive)} = \frac{0.95 \times 0.001}{0.05 times 0.999} = 0.019\\ P(disease \mid positive) + P(\neg disease \mid positive) = 1$$
Very low! we cannot compensate too much the prior even if the test is positive, only from 0.1% to about 2%, still far from 50%.