Events |
---|
- Both the sample space Ω and the empty set ∅ are in Σ.
- If a finite or countable number of events ω1, &omega2, ... belong to Σ, then their union and their intersection belongs to Σ.
- If event ω belongs to Σ, then its complement ω' belongs to Σ.
(1, 1) | (1, 2) | (1, 3) | (1, 4) | (1, 5) | (1, 6) |
(2, 1) | (2, 2) | (2, 3) | (2, 4) | (2, 5) | (2, 6) |
(3, 1) | (3, 2) | (3, 3) | (3, 4) | (3, 5) | (3, 6) |
(4, 1) | (4, 2) | (4, 3) | (4, 4) | (4, 5) | (4, 6) |
(5, 1) | (5, 2) | (5, 3) | (5, 4) | (5, 5) | (5, 6) |
(6, 1) | (6, 2) | (6, 3) | (6, 4) | (6, 5) | (6, 6) |
- The probability of an event ω ∈ Σ is a real nonnegative number: Pr(ω) ≥ 0.
- Pr[Ω] = 1.
-
If ω1, ω2, ... is a list of disjoint events
from Σ, then
\[ \Pr \left[ \omega_1 \cup \omega_2 \cup \omega_3 \cdots \right] = \Pr [\omega_1 ] + \Pr\left[ \omega_2 \right] + \Pr\left[ \omega_3 \right] + \cdots . \]This holds for finite or countable unions.
The third axiom is probably the most interesting one. The basic idea is that if some events are disjoint (i.e., there is no overlap between them), then the probability of their union must be the summations of their probabilities. Another way to think about this is to imagine the probability of a set as the area of that set in the Venn diagram. If several sets are disjoint such as the ones shown
Notice that the events that {A wins}, {B wins}, {C wins}, and {D wins} are disjoint since more than one of them cannot occur at the same time. For example, if A wins, then B cannot win. From the third axiom of probability, the probability of the union of two disjoint events is the summation of individual probabilities. Therefore,
Borel sets and σ-fields
Later on, we will discuss probabilities for sets from real axis ℝ, and we need to determine what sets from it can have probabilities because not every subset has it. Obviously, we need to have probability for every interval, and also for any set that is a union of finitely many intervals. For technical reasons, not every set can be assigned a probability, but these sets that can be built from intervals should be included.When one is trying to assign probabilities to every set from ℝ, he or she will get into troubles of two kinds. First, assigning probabilities to the subintervals does not uniquely determine the probabilities of all the other subsets. We would have to describe the assignment in some way for other sets; yet it is impossible even to describe all of these sets, not to mention deciding on their probabilities. Second, even if we could assign probabilities to the rest of the subsets in some way, our assignment would be inconsistent, in that there will be some countable collection of disjoint sets whose union has a probability that is not the sum of the individual sets probabilities. There is a famous Banach--Kuratowski theorem according to which no consistent assignment of probabilities is possible if we require that singletons have probability 0.
It is reasonable to include into the set having probabilities the following subsets as events: intervals, their unions, intersections, and complements. For all such subsets, we know how to define their probabilities. It is important to understand that we use not arbitrary unions or intersections, but only countable collection of events from which we are able to build new events. This will guarantee that probabilities of all events are determined uniquely once we know probabilities of subintervals from which they are constructed. Such subsets are called the Borel subsets named after Émile Borel (1871--1956). A family of sets that is closed under these three set operations---complementation, and countable unions and intersections--is called a σ-field or σ-algebra.
- Bartoszynski, T. and Halbeisen, L., On a theorem of Banach and Kuratowski and K-Lusin sets, Rocky Mountain Journal of Mathematics, 2003, Vol. 33, Number 4, pp. 1223--1231.
- Banach, S. and Kuratowski, K., Sur une generalisation du probleme de la mesure, Fundamenta Mathematicae, 1929, Vol. 14, pp. 127--131.