AXIOMATIC PROBABILITY AND POINT SETS

The axioms of Kolmogorov. Let S denote a sample space with a probability

measure P defined over it, such that probability of any event A ⊂ S is given by
P (A). Then, the probability measure obeys the following axioms:
(1) P (A) ≥ 0,
(2) P (S) = 1,
(3) If {A1 , A2 , . . . Aj , . . .} is a sequence of mutually exclusive events such that
Ai ∩ Aj = ∅ for all i, j, then P (A1 ∪ A2 ∪ · · · ∪ Aj ∪ · · ·) = P (A1 ) + P (A2 ) +
· · · + P (Aj ) + · · ·.
The axioms are supplemented by two definitions:
(4) The conditional probability of A given B is defined by

P (A ∩ B)
P (A|B) = ,
P (B)

(4) The events A, B are said to be statistically independent if

P (A ∩ B) = P (A)P (B).

This set of axioms was provided by Kolmogorov in 1936.

Operations on Sets. The axioms of probability concern sets of events. In order
to employ these axioms, it is necessary to invoke the rules of Boolean algebra,
which are associated with a pair of binary operations. First, we must define these
operations together with some special sets.
A binary operation of union, denoted by the symbol ∪, may be defined
relative to any two sets A and B. The operation generates the set

A ∪ B = {x; x ∈ A or x ∈ B}.
Here the word “or” is used in the inclusive sense to imply that x is either in A or in
B or in both. For example, if S is the set of all vertebrates, A is the characteristic
of having fur and B is the characteristic of laying eggs, then A ∪ B certainly has
the duck-bill platypus amongst its elements as well as foxes and geese.
A binary operation of intersection, denoted by the symbol ∩, may be de-
fined relative to any two sets A and B. The operation generates the set

A ∩ B = {x; x ∈ A and x ∈ B}.

In terms of the previous example, A ∩ B (unless I am mistaken) has only the duck-
bill platypus and the spiny ant eater as its two elements.
Two sets A and B are said to be disjoint if their intersection is the empty
set A ∩ B = ∅.

1
If A is the set of vertebrate fish and B is the set of mammals, then, according to
modern usage, their intersection is the empty set. However, as recently as Victorian
times, whales, which are mammals, were liable to be described as fish.
Let A ⊂ S. Then the complement of A in S, denoted by Ac , is the set of
all the elements of S which do not belong to A: Ac = {x; x ∈
/ A}.
The rules of Boolean Algebra. The binary operations of union ∪ and intersec-
tion ∩ are roughly analogous, respectively, to the arithmetic operations of addition
+ and multiplication ×, and they obey a similar set of laws. In fact, the laws of
Boolean algebra are virtually symmetric with respect to the two operations in the
sense that, in any of the statements of the laws that are listed below, the sym-
bols can be interchanged without affecting their truth. This in not the case in
arithmetic. The laws are as follows:
Commutative law: A ∪ B = B ∪ A,
A ∩ B = B ∩ A,
Associative law: (A ∪ B) ∪ C = A ∪ (B ∪ C),
(A ∩ B) ∩ C = A ∩ (B ∩ C),
Distributive law: A ∩ (B ∪ C) = (A ∩ B) ∪ (A ∩ C),
A ∪ (B ∩ C) = (A ∪ B) ∩ (A ∪ C),
Idempotency law: A ∪ A = A,
A ∩ A = A.
These various laws have the status of axioms.
There are several useful identities which are deducible from the axioms. Thus
De Morgan’s Rules state that

(A ∪ B)c = Ac ∩ B c and (A ∩ B)c = Ac ∪ B c .

Amongst other useful results are the following:

(i) A ∪ Ac = S, (iv) A ∩ S = A,
(ii) A ∩ Ac = ∅, (v) A ∪ ∅ = A,
(iii) A ∪ S = S, (vi) A ∩ ∅ = ∅.

Theorem: the union of of events. The probability that either A or B will happen
or that both will happen is the probability of A happening plus the probability of B
happening less the probability of the joint occurrence of A and B:

P (A ∪ B) = P (A) + P (B) − P (A ∩ B)

Proof. A ∪ B = A ∪ (B ∩ Ac ) is the union of two disjoint sets; so, according to

axiom 3, there is
P (A ∪ B) = P (A) + P (B ∩ Ac ).

2
But B = (B ∩ A) ∪ (B ∩ Ac ) is also the union of two disjoint sets, so there is also

P (B) = P (B ∩ A) + P (B ∩ Ac ) =⇒ P (B ∩ Ac ) = P (B) − P (B ∩ A).

Substituting the latter expression into the one above gives

P (A ∪ B) = P (A) + P (B) − P (A ∩ B).

Lemma: the probability of the null event. Axiom 3 implies that P (S ∪ ∅) =

P (S) + P (∅), since S and ∅ are disjoint sets by definition, i.e. S ∩ ∅ = ∅. But also
S ∪ ∅ = S, so P (S ∪ ∅) = P (S) = 1, where the second equality is from axiom 2.
Therefore, P (S ∪ ∅) = P (S) + P (∅) = P (S), so P (∅) = 0.

Bayes’ Theorem
Observe that the formula for conditional probability implies that

P (A ∩ B) = P (A|B)P (B) = P (B|A)P (A),

whence we have
P (B|A)P (A)
P (A|B) = .
P (B)
This is the basis of Bayes’ law of inverse probabilities, which provides an idealised
model of a process by which we might adapt our our probabilistic beliefs or hy-
potheses concerning the state of the world in view of the empirical evidence that
accumulates.
Consider a set Ω = {H1 , H2 , . . . , Hn }, wherein Hi ∩ Hj = ∅; i = j, which
comprises all possible explanations of an event E. Under some circumstances, it is
possible to define a probability measure over Ω that indicates the relative likelihoods
of the alternative hypotheses therein. In the absence of the evidence of E, they are
described as prior likelihoods or probabilities.
The evidence given by the event E will cast some light upon the likelihoods
of the hypotheses, which is to say that we can define a set of modified posterior
likelihoods over the set Ω. The posterior likelihood of an hypothesis Hi in the light
of the event E is given by

P (E|Hi )P (Hi )
P (Hi |E) = , where
P (E)
 
P (E) = P (E ∩ Hi ) = P (E|Hi )P (Hi ).
i i

We use the following terminology:

P (Hi ) is the prior likelihood of the ith hypothesis Hi ,
P (Hi |E) is the posterior likelihood of the ith hypothesis,
P (E|Hi ) is the conditional probability of the event E under the hypothesis Hi ,

3
 P (E) is the unconditional probability of the event E.
Example. The Manager of Ffyfes, who import bananas to the U.K. from many
sources, has discovered an unmarked crate, and he wishes to determine its origin.
40 percent of the crates in stock come from Guatemala and 60 percent from Cuba.
On average, 1/2 the Guatemalan bananas are bad and 1/6 of the Cuban bananas
are bad. The manager opens the crate and pulls out a banana that happens to be
bad. In the light of this evidence, what it is the most likely origin of the crate?
Answer. Let H1 denote the hypothesis that the crate if from Cuba and let H2
denote the hypothesis that it is from Guatemala. Let E be the event of discovering
a rotten banana. There are the following items of information:

60 3 20 2
P (H1 ) = = P (H2 ) = = ,
100 5 100 5
1 1
P (E|H1 ) = P (E|H2 ) = ,
6 2
1 3 1
P (E|H1 )P (H1 ) = × = P (H1 |E),
6 5 10
1 2 1
P (E|H2 )P (H2 ) = × = P (H2 |E).
2 5 5
In the light of the evidence, it seems twice as likely that Guatemala is where the
crate is from.