You are on page 1of 12

The Language of Probability

The "universe" consists of all the "elementary events," precisely one of

which will actually occur.

For example, a five-card hand is dealt from a deck of 52 cards. There are
2,598,960 different hands which might be dealt, and the dealing of any
one of them could be considered an elementary event.

To each elementary event is attached a "probability," which represents the universe (all possible wo
the relative likelihood of that occurrence. These probabilities are all
non-negative, and sum to 1. In the example, assuming a fair deal, it
makes sense to assign equal probabilities to all possible deals, so each
has a probability of 1/2,598,960.

an elementary event
e.g. [2,3,7,9,Q]

N.B. The "counts" on this and the following tabs are derived in
the "Counting" workbook.
e universe (all possible worlds)

An "event" is a set of elementary events, and is said to "happen" if the

actual elementary event which occurs is in that set.

The probability of an event is the sum of the probabilities of the elementary

events which would lead to the event happening.

In our five-card example, let A represent the event "the hand contains at
least one ace." the event A
"A happens"
There are 886,656 hands containing at least one ace. Therefore,

Pr(A) = 886,656 / 2,598,960 = 0.3412 = 34.12%

If A is any event, the complement of A (symbolically, A c) is the event that

A doesn't happen, i.e., that the elementary event which occurs does not
lie in the set A.

Clearly, Pr(A) + Pr(Ac) = 1 . Either A happens, or it doesn't.

the event Ac
'"A doesn't happen"
shorter: "not A"

the event A
"A happens"
Let B represent the event "the hand contains no clubs." Since there are
575,757 hands containing no clubs,

Pr(B) = 575,757 / 2,598,960 = 0.2215 = 22.15%

The event "A and B both happen" is represented symbolically as AB

(the intersection of the set of elementary events constituting A, and the AB
set constituting B). A-B = ABc
"A and B"
There are 198,765 hands consisting of no clubs, and containing at least "A and not B"
one ace. Therefore

Pr(AB) = 198,765 / 2,598,960 = 0.0765 = 7.65%

The event "A happens, but B doesn't" is just AB c, which is sometimes AcBc "not A and not B
written as A-B.

Clearly, Pr(A) = Pr(AB) + Pr(ABc), and Pr(B) = Pr(BA) + Pr(BAc) .

For example, the first relationship can be used to find the probability that
the dealt hand will contain at least one ace (A) and at least one club (B c).

Pr(ABc) = Pr(A) - Pr(AB) = 26.47%


B-A = BAc
"A and B"
"B and not A"

AcBc "not A and not B"

The event that either A or B (or both - the "or" is inclusive) happens is
denoted by AUB. If we add together all cases where A happens, and all
cases where B happens, then the cases where both A and B happen get
counted twice. Therefore,

Pr(AUB) = Pr(A) + Pr(B) - Pr(AB) .

As well, by comparing this chart with the previous one, we see that
"Either A, or B, or both"
(AUB)c = AcBc . shorter: "A or B"

Therefore, since Pr( (AUB)c ) = 1 - Pr(AUB),

Pr(AUB) = 1 - Pr(AcBc) .
(AUB)c "neither A nor B


"Either A, or B, or both"

shorter: "A or B"

(AUB)c "neither A nor B"


Examining the diagram to the right, we see seven events, any one of
which might be of interest.
Browsing back over the last two tabs, we find four relationships tying
the probabilities of these events together:

Pr(A) = Pr(AB) + Pr(ABc)

Pr(B) = Pr(BA) + Pr(BAc)

Pr(AUB) = Pr(A) + Pr(B) - Pr(AB)

Pr(AUB) = 1 - Pr(AcBc)
Why are these relationships of interest? Because, given the
probabilities of any three of the seven events, these four relationships
together determine all the other probabilities!

If you're trying to compute the probability of an event, and can't see how
to do the computation directly, these four "rules" (really, three rules, since
the first two involve the same idea) can help you to reformulate your
problem in a manner which makes the calculations easier.



Most fatal automobile accidents involve a drunk driver, or a driver who

was exceeding the legal speed limit. Of course, a number of these
accidents involve drunk drivers and speeding. 68%

Assume that you already know that 68% of all fatal automobile accidents
involve drunk drivers (and possibly speeding as well, but possibly not),
55% involve speeding (and possibly a drunk driver, but possibly not), and
43% involve drunk drivers and speeding.

Consider a particular fatal accident, selected at random. If A = "the

accident involved a drunk driver," and B = "the accident involved
speeding," then what we know is

Pr(A) = 68% Pr(B) = 55% Pr(AB) = 43%

Recall the four rules from the previous page:

Pr(A) = Pr(AB) + Pr(ABc) 68% = 43% + Pr(ABc)

Pr(B) = Pr(BA) + Pr(BAc) 55% = 43% + Pr(BAc)

which become
Pr(AUB) = Pr(A) + Pr(B) - Pr(AB) Pr(AUB) = 68% + 55% - 43

Pr(AUB) = 1 - Pr(AcBc) 80% = 1 - Pr(AcBc)

Therefore, for example, the probability that the accident involved neither
a drunk driver nor speeding is 20%. And the probability that it involved
speeding, but no drunk driver, is only 12%.

Do notice that, while the rules can be applied "algebraically," their

application in the context of the diagram is quite simple.

Try this problem:

You have collected some demographic data on members of an organization

for which you are preparing a conference. In particular, you've noted the sex
and dining preference of each conference registrant.

Unfortunately, you've just spilled coffee on your notes! All you can still make
out is that 60% are male, 30% are female and want vegetarian meals, and
a total of 55% want meat platters (rather than vegetarian platters).
What fraction of the group consists of men wanting meat platters? 45%
What fraction are women wanting meat platters? 10%


68% = 43% + Pr(ABc)

55% = 43% + Pr(BAc)

Pr(AUB) = 68% + 55% - 43% = 80%

80% = 1 - Pr(AcBc)