Mutually Exclusive Events

Deep Breakdown

What it tells me (Situation Awareness): Two events are mutually exclusive if they cannot both occur at the same time: $P(A \cap B) = 0$. When this holds, the union probability simplifies to the sum of individual probabilities. This is Kolmogorov's additivity axiom. In practice, it applies to exhaustive category memberships: a customer cannot be in both "churned" and "active" simultaneously.

Flowchart

Real-world use

Where it is used

Multi-class classification: classes are mutually exclusive (a data point belongs to exactly one class). Also: customer state transitions (active/churned/paused), product category assignment, or diagnostic test outcomes.

Problem it solves

When events are mutually exclusive, the union rule simplifies to addition — no intersection term needed.

How it solves the problem

Since $P(A \cap B) = 0$, the inclusion-exclusion formula $P(A \cup B) = P(A) + P(B) - P(A \cap B)$ reduces to $P(A) + P(B)$.

Example in a business

Multinomial classification with 3 sentiment classes: P(positive) = 0.55, P(neutral) = 0.30, P(negative) = 0.15. Classes are mutually exclusive, so $P(\text{pos or neu}) = 0.55 + 0.30 = 0.85$.

The formula, piece by piece

$P(A \cap B) = 0 \implies P(A \cup B) = P(A) + P(B)$

• $P(A \cap B) = 0$: A and B have no overlap.
• $P(A \cup B) = P(A) + P(B)$: the additive rule.

The formula in words

"If A and B cannot happen at the same time (mutually exclusive), the probability that A OR B occurs is simply the sum of their individual probabilities."

Critical Concepts

• Mutual exclusivity does NOT imply independence (and vice versa).
• In fact, if $P(A) > 0$ and $P(B) > 0$ and they're mutually exclusive, they CANNOT be independent.
• Exhaustive + mutually exclusive = a partition: $\sum_i P(A_i) = 1$.
• Categorical distribution uses this: one class per sample.

Step-by-step calculation (tiny example)

Step 1: Customer states A: churned ($P = 0.20$), B: paused ($P = 0.10$). States are mutually exclusive.

Step 2: P(churned OR paused) $P(A \cup B) = 0.20 + 0.10 = 0.30$

Step 3: Verify no overlap No customer can be both churned and paused simultaneously. $P(A \cap B) = 0$. Correct.

Code examples

Beginner example (Python)

# Multi-class classification: mutually exclusive classes
class_probs = {"positive": 0.55, "neutral": 0.30, "negative": 0.15}   # P(positive or neutral)
p_pos_or_neu = class_probs["positive"] + class_probs["neutral"]
print(f"P(positive or neutral) = {p_pos_or_neu:.2f}")   # 0.85

Beginner example (R)

class_probs <- c(positive=0.55, neutral=0.30, negative=0.15)
p_pos_or_neu <- class_probs["positive"] + class_probs["neutral"]
cat("P(positive or neutral) =", p_pos_or_neu, "\n")  # 0.85

Professional example (Python)

import numpy as np
from typing import Sequence

def union_mutually_exclusive( probabilities: Sequence[float], event_names: Sequence[str] | None = None ) -> float:
    """Compute P(union) for mutually exclusive events via simple addition. Validates mutual exclusivity assumption (all probs fit within [0,1] and total does not exceed 1). Args: probabilities: Individual P(Ai) values. event_names: Optional names for reporting. Returns: P(A1 ∪ A2 ∪ ... ∪ Ak) = sum of P(Ai). """
    prob_array = np.array(probabilities)
    if np.any(prob_array < 0):
        raise ValueError("Negative probability found")
    total_union = float(np.sum(prob_array))
    if total_union > 1.0 + 1e-8:
        raise ValueError(f"Sum {total_union:.6f} > 1: not mutually exclusive or invalid probs")
    return total_union   # Sentiment class union analysis
sentiment_probs = [0.55, 0.30, 0.15]
event_names = ["positive", "neutral", "negative"]
for k in range(1, len(sentiment_probs) + 1):
    p_union_k = union_mutually_exclusive(sentiment_probs[:k], event_names[:k])
names_str = " ∪ ".join(event_names[:k])
print(f"P({names_str}) = {p_union_k:.2f}")

Source: AIO p. 84

#probability-statistics#probability-theory#axioms-and-basic-probability