0% found this document useful (0 votes)
2 views3 pages

Bayes Theorem

The document discusses the application of probability and Bayes' theorem in determining the likelihood of genetic traits and medical diagnoses. It illustrates how new information, such as the birth of an unaffected child, can change the probability of future outcomes for genetic traits. Additionally, it presents a case study on the misdiagnosis of AIDS using Bayes' theorem, highlighting the importance of understanding conditional probabilities in medical testing.

Uploaded by

balamicro47
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
2 views3 pages

Bayes Theorem

The document discusses the application of probability and Bayes' theorem in determining the likelihood of genetic traits and medical diagnoses. It illustrates how new information, such as the birth of an unaffected child, can change the probability of future outcomes for genetic traits. Additionally, it presents a case study on the misdiagnosis of AIDS using Bayes' theorem, highlighting the importance of understanding conditional probabilities in medical testing.

Uploaded by

balamicro47
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

Lecture 5

Consider the following pedigree of an autosomal recessive trait.

= female

= male

p(affected child) = p(mother carrier and father carrier and affected child)

= 2/3 x 2/3 x 1/4 = 1/9

However, if they have a child that is affected we must reassess the probability that
their next child will be affected.

p(both parents carriers) = 1. So, p(next child affected) = 1/4

This example shows how probability calculations are based on information. The
probability changes not because the parents have changed but because our information
about them has.

Now consider the case that the two parents have an unaffected child, with this new
information we can recalculate the probability that the next child will be affected.
An unaffected child does not establish definitively that both parents are not carriers,
but it should be apparent that our estimate of the probability should be somewhat less
than the original probability of 4/9.

In order to calculate such probabilities we need to introduce a new concept known as


conditional probability.

p(X|Y) = probability that event X will occur given that Y has occurred

p(X and Y)
p(X|Y) =
p(Y)

Therefore:

p(X and Y) = p(X|Y) • p(Y)


Bayes Theorem

We can use this identity to perform a very useful type of probability calculation. In many
simple probability calculations we can readily calculate the probability of a particular
outcome (effect) given a particular circumstance (cause). Bayes’ theorem allows us to
calculate the inverse probability of a particular cause based on an observed effect. This
method depends on knowing the probabilities of the measured event occurring given each
of the possible causes and on knowing the a priori probabilities of obtaining the
conditions for each of the possible causes.

Stated in terms of conditional probabilities, Bayes’ theorem allows us to express


p(X|Y) in terms of p(Y|X), where X represents a particular circumstance (cause) and Y
represents an observed effect.

p(X and Y) p(Y|X) • p(X)


p(X|Y) = =
p(Y) p(Y)

Consider a simple case in which there are only two possible circumstances/hypotheses
(ie both parents are carriers or both parents are not carriers). We will express the
complement of X (not X) as X. Thus:

p(Y) = [p(Y|X) • p(X)] + [p(Y|X) • p(X)]

p(Y|X) • p(X)
p(X|Y) =
[p(Y|X) • p(X)] + [p(Y|X) • p(X)]

To apply this formula to the pedigree problem we will define X = both parents carriers;
and X = not both parents carriers; Y = first child unaffected.

Accordingly, p(X) = 4/9, p(X) = 5/9, p(Y|X) = 3/4, p(Y|X) = 1

3/4 • 4/9 3/9


p(X|Y) = = = 3/8
(3/4 • 4/9) + (1 • 5/9) 3/9 + 5/9

The probability that the next child will be affected is 3/8 x 1/4 = 3/32 = 0.094, which
is slightly less than the probability we calculated before the couple had an unaffected
child 1/9 = 0.111
Here is an alternative way to express Bayes theorem, which shows how this expression fits
well with common sense.

p(Y|X) • p(X)
p(X|Y) = p(Y)

p(Y|X) • p(X)
p(X|Y) =
p(Y)

Thus: p(X|Y) p(Y|X) p(X)


= •
p(X|Y) p(Y|X) p(X)

Posterior = Likelihood • Prior


odds ratio odds

Let’s now consider two examples to show the power of Bayes’ theorem to analyze a wide
variety of interesting problems.

For example, lets caculate the probability of a misdiagnosis of AIDS infection using a
diagnostic test for HIV that has both a false positive and false negative rate = 0.005,
and given that the a priori probability that an individual in the US has AIDS is 0.001.
Expressing each of these terms as conditional probabilities we have:

p(Pos|Inf) = 0.995 (probability that an individual tests positive given they are infected)

p(Pos|NInf) = 0.005 (probability that an individual tests positive given they are not
infected)

Probability that an individual in the US is infected p(Inf) = 0.001

Using Bayes’ theorem:


p(Pos | Inf) • p(Inf)
p(Inf | Pos) =
p(Pos | Inf) • p(Inf) + p(Pos | NInf) • p(NInf)

= 0.995 • 0.001
(0.995 • 0.001) + (0.005 • 0.999)

= 0.16

The remarkable conclusion is that although the AIDS test has a very low error rate (for
individuals who test positive there is a 99.5% chance that they have the disease) when the
test is used broadly, only a minority of the positive tests would actually be AIDS cases.

You might also like