1 Sample space and event definitions
1.1 Probability experiments and outcomes
A probability experiment is a procedure whose result is uncertain prior to observation. The outcomes are the basic results the experiment can produce. Collectively, they form the sample space, often denoted by \(S\). The sample space may be finite, countably infinite, or uncountably infinite depending on the experiment.
Outcomes are treated as the finest level of description used in the model. Probabilities are then assigned to events, where an event groups outcomes according to a rule.
1.2 Events as subsets of the sample space
An event is any collection of outcomes from the sample space that can be described in the model. In set-theoretic probability, an event corresponds to a subset \(E \subseteq S\). This viewpoint supports operations such as union, intersection, and complement, which correspond to “at least one of,” “both,” and “none of,” respectively.
Not all subsets may be allowable in advanced settings (for example, measure-theoretic probability), but in standard discrete examples, treating events as all subsets of \(S\) is common.
1.3 Exhaustive vs. non-exhaustive collections
A collection of events is exhaustive if it collectively accounts for every outcome in the sample space. Intuitively, the union of the events includes all possible outcomes: \[ E_1, E_2, \dots, E_n \text{ is exhaustive if } E_1 \cup E_2 \cup \cdots \cup E_n = S. \] If the union is a proper subset of \(S\), then the collection is non-exhaustive, meaning some outcomes are not included in any event from the list.
This distinction matters because it determines whether probabilities over the collection capture the full probability mass of the experiment.
2 Exhaustive events in probability theory
2.1 Formal definition of exhaustiveness
2.1.1 Exhaustive coverings of the sample space
Consider a finite or countable family of events \(\{E_i\}_{i \in I}\) within a sample space \(S\). The family is exhaustive if every outcome lies in at least one event: \[ \bigcup_{i\in I} E_i = S. \] This is a covering property: the events “cover” the sample space without necessarily being disjoint.
2.1.1.1 “At least one occurs” interpretation
Exhaustiveness can be read in terms of occurrences: for each trial, at least one event in the collection happens. If a particular trial produces outcome \(\omega\), then there must exist an index \(i\) such that \(\omega \in E_i\). Otherwise, \(\omega\) would fall outside the union and the collection would not be exhaustive.
2.2 Probability statement for exhaustive sets
2.2.1 Relation to total probability
If \(\{E_i\}\) is exhaustive, then the probability that at least one of them occurs equals 1: \[ P\left(\bigcup_{i} E_i\right) = P(S) = 1. \] This identity holds regardless of whether the events overlap.
A familiar special case arises when the events are also mutually exclusive. Then the probability of the union equals the sum of probabilities, yielding a partition of total probability mass.
2.3 Constructing an exhaustive set
To construct an exhaustive set, a common strategy is to define events that describe a complete set of categories for outcomes. For discrete problems, one often selects events based on a range of values that span all possibilities. For example, when outcomes are outcomes of a die, events can be defined by the face value (such as “shows 1,” “shows 2,” etc.), and together they form an exhaustive collection.
Another approach is to start with a proposed list and then verify exhaustiveness by checking whether outcomes excluded from the list have probability zero (or, in a purely set-theoretic sense, whether the complement is empty).
3 Exhaustive and mutually exclusive event sets
3.1 Mutually exclusive events
Two events \(A\) and \(B\) are mutually exclusive if they cannot both occur in the same trial, which in set terms means: \[ A \cap B = \varnothing. \] For multiple events \(\{E_i\}\), mutual exclusivity typically means that no outcome belongs to two different events in the family: \[ E_i \cap E_j = \varnothing \quad \text{for } i \neq j. \]
3.2 Partition of the sample space
3.2.1 When the sum of probabilities equals 1
When a collection of events is both exhaustive and mutually exclusive, it forms a partition of the sample space. In this case, the union is the whole sample space and the events do not overlap. The probability then satisfies: \[ P(E_1) + P(E_2) + \cdots + P(E_n) = 1. \] This follows from additivity over disjoint events: \[ P\left(\bigcup_{i=1}^n E_i\right) = \sum_{i=1}^n P(E_i), \] together with \( \bigcup_{i=1}^n E_i = S\).
3.3 Common partition examples
Partitions frequently arise by classifying outcomes according to mutually exclusive categories. Common constructions include:
- Outcome-by-value events (e.g., “die shows 1,” “die shows 2,” and so on).
- Sign or parity categories (e.g., “even” vs. “odd” in a setting where all outcomes are covered).
- Ranges that cover all allowed values without overlap.
Care is needed: categories can be exhaustive but not mutually exclusive (overlap), or mutually exclusive but not exhaustive (missing outcomes). The partition property requires both.
4 Examples and worked illustrations
4.1 Dice outcomes
Let the experiment be a single roll of a fair six-sided die. The sample space is \[ S=\{1,2,3,4,5,6\}. \] Define events: \[ E_1=\{\text{roll is even}\}=\{2,4,6\},\quad E_2=\{\text{roll is odd}\}=\{1,3,5\}. \] The events are mutually exclusive because no number can be both even and odd. They are exhaustive because every face value is either even or odd. Therefore, \[ P(E_1)+P(E_2)=1. \] For a fair die, \(P(E_1)=3/6\) and \(P(E_2)=3/6\), so the sum is \(1\).
4.2 Coin toss sequences
Consider two coin tosses with outcomes represented as ordered pairs: \(S=\{HH,HT,TH,TT\}\). Define: \[ E_1=\{\text{exactly one head}\}=\{HT,TH\},\quad E_2=\{\text{no heads}\}=\{TT\},\quad E_3=\{\text{two heads}\}=\{HH\}. \] These events are mutually exclusive: an outcome cannot have both zero heads and one head. They are also exhaustive: every two-toss sequence has either 0, 1, or 2 heads. Hence, \[ P(E_1)+P(E_2)+P(E_3)=1. \] With a fair coin, each of the four outcomes has probability \(1/4\), giving \(P(E_1)=2/4\), \(P(E_2)=1/4\), and \(P(E_3)=1/4\), whose sum equals \(1\).
4.3 Cards and simple draws
Suppose one card is drawn uniformly from a standard 52-card deck. Let \(S\) be the set of all 52 cards. One can partition \(S\) using suits: \[ E_{\clubsuit}=\{\text{card is a club}\},\quad E_{\spadesuit}=\{\text{card is a spade}\},\quad E_{\heartsuit}=\{\text{card is a heart}\},\quad E_{\diamondsuit}=\{\text{card is a diamond}\}. \] Each event has 13 cards, the events are disjoint, and their union is all 52 cards. Thus, \[ P(E_{\clubsuit})+P(E_{\spadesuit})+P(E_{\heartsuit})+P(E_{\diamondsuit})=1. \]
Alternatively, one can define an exhaustive collection using ranks, such as events “card is an Ace,” “card is a King,” etc. There will be 13 ranks, each with four cards, again producing a partition.
4.4 Urn models and labeled objects
Consider an urn containing labeled objects. Suppose there are 5 red balls labeled \(R_1,\dots,R_5\) and 3 blue balls labeled \(B_1,\dots,B_3\). Draw one ball uniformly at random. The sample space can be taken as the eight labeled balls: \[ S=\{R_1,\dots,R_5,B_1,\dots,B_3\}. \] Define: \[ E_R=\{\text{draw is red}\}=\{R_1,\dots,R_5\},\quad E_B=\{\text{draw is blue}\}=\{B_1,\dots,B_3\}. \] Then \(E_R\) and \(E_B\) are mutually exclusive and exhaustive. Therefore, \[ P(E_R)+P(E_B)=1, \] and numerically \(P(E_R)=5/8\), \(P(E_B)=3/8\).
If instead one defined events like \(E_{R_1}=\{R_1\}\), \(E_{R_2}=\{R_2\}\), etc., those singleton events also form a partition: every possible labeled outcome lies in exactly one singleton event.
5 Checking whether a set is exhaustive
5.1 Coverage tests using complements
A direct way to test exhaustiveness uses complements. A collection \(\{E_i\}\) is exhaustive if and only if the complement of their union is empty: \[ \left(\bigcup_i E_i\right)^c=\varnothing. \] Equivalently, \[ \bigcup_i E_i = S \quad \Longleftrightarrow \quad S \setminus \bigcup_i E_i = \varnothing. \] In words, there should be no outcome that lies outside every event in the collection.
5.2 Graphical and table-based verification
For discrete models, exhaustiveness can be checked by listing outcomes in a table and marking which events contain each outcome. For instance:
- In a dice problem, one can list faces 1 through 6 and verify each face appears in at least one event column.
- In coin-toss sequences, one can enumerate the four ordered pairs and confirm each pair belongs to at least one event.
Graphical methods, such as Venn diagrams, also help when the number of events is small. The crucial feature is whether the union covers the entire sample space region.
5.3 Edge cases and improper event descriptions
Some sets may fail exhaustiveness due to boundary omissions. Examples include:
- Ranges that do not cover all possible values (e.g., “greater than 0” and “less than 0” but omitting “equals 0”).
- Event definitions that implicitly assume conditions not stated in the sample space.
- Overlapping events that are still exhaustive (overlap alone does not break exhaustiveness), which can confuse attempts to interpret “coverage” as “exclusivity.”
In addition, an empty event contributes nothing to the union and cannot help achieve exhaustiveness. When using informal descriptions, it is helpful to translate each event into a precise subset of \(S\).
6 Connections to other probability concepts
6.1 Complements and “what’s left”
The complement perspective is closely related to exhaustiveness. If \(\{E_i\}\) is not exhaustive, then there exists a “leftover” event consisting of outcomes in none of the \(E_i\). This leftover set is: \[ L = S \setminus \bigcup_i E_i. \] Then \[ P(L)=1-P\left(\bigcup_i E_i\right). \] Thus, exhaustiveness corresponds to the leftover probability being zero (and in strict set terms, the leftover set being empty).
6.2 Conditional probability with exhaustive cases
Exhaustive partitions provide a natural setting for conditional probability. Suppose \(\{E_i\}\) is a partition of \(S\). Then, for any event \(A\), \[ P(A)=\sum_i P(A\cap E_i). \] Because \(E_i\) disjointly cover \(S\), \(A\cap E_i\) isolates the portion of \(A\) that occurs alongside each case \(E_i\).
This structure is useful when one wants to analyze how an outcome of interest \(A\) distributes across mutually exclusive categories.
6.3 Bayes-style reasoning via exhaustive partitions
Bayes-style arguments often rely on partitions that represent competing hypotheses. Let \(\{E_i\}\) be an exhaustive mutually exclusive set representing possible scenarios, and let \(A\) be observed evidence. Then conditional probabilities of hypotheses given evidence can be expressed using: \[ P(E_i \mid A)=\frac{P(A\mid E_i)P(E_i)}{\sum_j P(A\mid E_j)P(E_j)}. \] The denominator becomes a total-probability sum precisely because the \(E_j\) form an exhaustive partition. This ensures that all scenarios contributing to the event \(A\) are accounted for.