28 Days Earlier

In a city with $N = 100,000$ inhabitants $N P(I) = 100$ persons are infected. Out of them $N P(T|I) P(I) = 99$ persons are tested positive. Out of the healthy population of $N P(\overline{I}) = 99,900$ people a percentage of $P(T|\overline{I}) = 1 - P(\overline{T}|\overline{I}) = 2\,\text{\%}$ are also tested positive, so that we have $N P(T|\overline{I}) P(\overline{I}) = 1998$ false positive test results. The probablitity for a positive tested person to be infected results $P(I|T) = \frac{P(T|I) P(I)}{P(T|I) P(I) + P(T|\overline{I}) P(\overline{I})} = \frac{99}{99 + 1998} = 4.721\,\text{\%} \approx 5 \,\text{\%}$ Altough the test has a fairly large sensitivity and specitivity, only $5 \,\text{\%}$ of the positive tests are actually correct due to the small amount of infections in the whole population.

For the second test we have group of $N' = 2097$ people with $99$ infections. Out of them $98$ persons are tested a second time true positive. Out of $1998$ healthy people there are still $40$ false positive test. Therefore, the infection probability results $P(I|T_1 \cap T_2) \approx \frac{98}{98 + 40} \approx 71\,\text{\%}$ After the second positive test result, Paul is probably infected, but we are still far from certainty.