False positive

A new disease has been identified in the human population. For the sake of the following discussion, we will refer to it as C 1 C1 (consumerism-order 1).

Scientists have have devised a machine for testing a person for C 1 C1 . It identifies C 1 C1 in an infected person with 99 % 99\% accuracy. It identifies a healthy person as healthy 95 % 95\% of the time. The probability of a random person being affected by C 1 C1 is known to be 1 0 4 10^{-4} .

Find the approximate probability that a person is affected by C 1 C1 , given that his/her test returned positive.

Note: The data given in this problem are representative. The actual probability of a random human to be affected by C 1 C1 is best left unsaid.

0.002 0.998 0.999 all other options are incorrect 0.001

This section requires Javascript.
You are seeing this because something didn't load right. We suggest you, (a) try refreshing the page, (b) enabling javascript if it is disabled on your browser and, finally, (c) loading the non-javascript version of this page . We're sorry about the hassle.

3 solutions

Peter Macgregor
Apr 26, 2015

I will use C to mean that a randomly selected individual is infected, C ˉ \bar{\text{C}} to mean that they do not have the infection and "positive" to mean that the test indicates that the person has the disease.

I will use P(A|B) to mean the probability that A is true given that B is true. So we want to find P(C|pos)

We can use the multiplicative rule for conditional probabilities to find P(C and pos) in two ways

P(C and pos) = P(C) P(pos|C) = P(pos) P(C|pos)

and so

P ( C p o s ) = P ( p o s C ) P ( C ) P ( p o s ) ( 1 ) P(C|pos)=\dfrac{P(pos|C)P(C)}{P(pos)}\dots(1)

The denominator can be written as the sum of the probabilities of two mutually exclusive possibilities - either the selected individual is infected and the test gives the correct result, or the person is not infected and the test gives a so called false positive. So (1) can be written as

P ( C p o s ) = P ( p o s C ) P ( C ) P ( C ) P ( p o s C ) + P ( C ˉ ) P ( p o s C ˉ ) P(C|pos)=\dfrac{P(pos|C)P(C)}{P(C)P(pos|C)+P(\bar{C})P(pos|\bar{C})}

Finally put in the numbers from the question to get

P ( C p o s ) = 0.99 × 1 0 4 0.99 × 1 0 4 + 0.05 × 0.999 0.002 P(C|pos)=\dfrac{0.99\times10^{-4}}{0.99\times 10^{-4}+0.05\times 0.999}\approx \boxed{0.002}

Assume we test 1 0 6 10^6 people.

Hence, 99 99 people are tested true positive and 49995 49995 are tested false positive.

Therefore, the required probability is: 99 49995 + 99 = 0.00198 0.002 \frac {99} {49995+99}= 0.00198\approx 0.002

This is an illustration of the False positive paradox

Got it wrong. Understood my mistake seeing the solution.

Anyways it was a nice question!

Sravanth C. - 6 years, 1 month ago

It should be (false positive+correct identification) in your denominator shouldn't it..?

99/(99+49995)=0.002

Rohit Sachdeva - 6 years, 1 month ago

Log in to reply

Edit made. Thanks

Raghav Vaidyanathan - 6 years, 1 month ago
Freddie Kalaitzis
Apr 26, 2015

p(negative | not affected ) = 0.95.

Thus, p(positive | not affected ) = 0.05.

p(positive & not affected) = p(positive | not affected) * p(not affected)

= 0.05 * 0.9999 = 0.049995.

p(positive & affected) = p(positive | affected) * p(affected)

= 0.99 * 0.0001 = 0.000099.

p(positive) = p(positive & (not affected or affected))

= 0.049995 + 0.000099 = 0.050094.

Bayes' rule:

p(affected | positive) = p( positive | affected ) * p( affected ) / p(positive)

= 0.99 * 0.0001 / 0.050094 = 0.00197628458

0 pending reports

×

Problem Loading...

Note Loading...

Set Loading...