Skip to content

Probability Mass Function

Represent a discrete random variable’s law by the nonnegative singleton probabilities p(x)=P(X=x), whose sum over its countable support equals one.

Version
v2 · 2026-09-06 · History
Domain-specific #
2541
Origin domain
mathematics
Subdomain
probability theory
Aliases
Pmf, Discrete probability function, Discrete density with respect to counting measure

Core Idea

For a discrete random variable \(X\), its probability mass function is

\[ p_X(x)=\Pr(X=x). \]

It assigns a nonnegative probability to each possible value, is positive on at most a countable support, and sums to one. For any event \(B\) in the value space, \(\Pr(X\in B)=\sum_{x\in B}p_X(x)\). Measure-theoretically, it is the Radon–Nikodym density of the law of \(X\) with respect to counting measure.[1]

The recognition invariant is discrete law + singleton probabilities + nonnegativity + countable summation to one.

Structural Signature

  • A probability space and a discrete random variable \(X\).
  • A finite or countable effective support.
  • Values \(p_X(x)=P(X=x)\).
  • Nonnegativity at every value.
  • Total mass \(\sum_xp_X(x)=1\).
  • Zero values outside the support when the function is extended to a larger codomain.
  • Event probabilities obtained by summing, not integrating against Lebesgue measure.
  • A pushforward probability law concentrated on atoms.
  • Density relative to counting measure.
  • Marginalization by summing a joint PMF.
  • Expectations computed as weighted sums.
  • Transformation by collecting masses over preimages.
  • Mode as an argument maximizing mass, possibly nonunique.

What It Is Not

It is not a continuous probability density function. A density value is not generally a point probability and may exceed one; a PMF value is itself a probability and lies in \([0,1]\). Continuous laws assign zero mass to each singleton even when they have a positive density there.[2]

It is not the random variable itself, nor the underlying sample-space probability measure. It is a function representing the random variable’s induced law at its atoms.

Scope of Application

PMFs describe counts, categories, finite-state outcomes, discrete-time events, combinatorial random objects, contingency tables, queue lengths, and digitized observations. Bernoulli, binomial, geometric, Poisson, categorical, and many empirical discrete laws are specified by PMFs.[3]

Countability is decisive. A mixed distribution can have atomic masses and a continuous component; listing the atom masses alone does not specify the entire mixed law.

Clarity

The support need not be a subset of the integers. It can be any finite or countable set whose singletons are measurable. If the variable is numerical, defining \(p_X\) on all real numbers with zeros off the support is convenient but not required.

The phrase “discrete density” is acceptable only when the reference measure is stated as counting measure. Calling a PMF a PDF without that qualifier invites the false rule that singleton probability must be obtained by integration over an interval.

Manages Complexity

A PMF compresses the full law into a lookup or formula from which events, moments, transforms, likelihoods, conditional probabilities, and simulations can be derived. Normalization and nonnegativity provide immediate validity checks.

For high-dimensional or large-support variables, direct tables become costly. Factorization, conditional independence, probability-generating functions, or sparse representations may manage the same mass without enumerating every state.

Abstract Reasoning

  1. Identify the random variable and its value space.
  2. Establish that its law is supported on a finite or countable set.
  3. Compute each singleton probability.
  4. Verify nonnegativity and total mass one.
  5. Define the support as values with positive mass.
  6. Sum masses for requested events.
  7. Obtain marginals by summing over unused coordinates.
  8. Compute expectations as \(\sum_xg(x)p_X(x)\) when the sum is defined.
  9. Treat mixed or continuous components separately.

Knowledge Transfer

The portable structure is a normalized nonnegative weight function over countable alternatives. The proposed immediate parent is Probability Distribution.

Examples

Fair die. \(p(x)=1/6\) for \(x\in\{1,\ldots,6\}\) and zero otherwise.

Poisson count. \(p(k)=e^{-\lambda}\lambda^k/k!\) for nonnegative integers \(k\).

Transformation. If \(Y=X^2\), then \(p_Y(y)\) is the sum of \(p_X(x)\) over every \(x\) with \(x^2=y\).

Non-example. The standard normal formula is a density, not a PMF.

Structural Tensions

  • Concrete tables versus symbolic formulas.
  • Singleton mass versus interval density.
  • Full support enumeration versus sparse factorization.
  • Exact normalization versus numerical rounding.
  • Finite laws versus countably infinite tails.
  • Atomic component versus complete mixed distribution.
  • Probability representation versus subjective probability weighting.

Structural–Framed Character

Normalized weighting, support, aggregation, and lookup are structural. Random variables, probability measures, atoms, counting measure, expectation, and statistical likelihood are probability frame.

Structural Core vs. Domain Accent

The portable core is distributing a conserved unit mass across countable alternatives. The constitutive domain accent is the probabilistic interpretation, induced law, singleton events, measure-theoretic reference, and sum-based calculus.

Probability Distribution is the proposed immediate parent. Probability, Function, Normalization, Support, Random Variable, and Expectation are related.

The prospective queue contains one strict edge to domain_specific:probability_distribution. No live DAG mutation is authorized.

Relationships to Other Abstractions

Local relationship map for Probability Mass FunctionParents appear above the current abstraction, mutual partners to the right, and children below. Node labels state whether each abstraction is prime or domain-specific; colors identify relation types.ProbabilityMass FunctionDOMAINDomain-specific abstraction: Probability Distribution — is a kind ofProbabilityDistributionDOMAIN

Current abstraction Probability Mass Function Domain-specific

Parents (1) — more general patterns this builds on

  • Probability Mass Function is a kind of Probability Distribution Domain-specific

    Probability Distribution is the proposed immediate parent.

Hierarchy paths (5) — routes to 3 parentless roots

Neighborhood in Abstraction Space

Probability Mass Function sits in a sparse region of the domain-specific corpus (89th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-09-08

Not to Be Confused With

  • Probability density with respect to Lebesgue measure.
  • Cumulative distribution function.
  • Probability weighting function in behavioral decision theory.
  • Frequency table without a probability model.
  • The random variable or sample space itself.
  • Only the atom list of a mixed distribution.
  • A likelihood viewed as a function of parameters.

References

[1] Patrick Billingsley, Probability and Measure, 3rd ed., Wiley, 1995. registry ↩

[2] NIST/SEMATECH, “What Is a Probability Distribution?”, e-Handbook of Statistical Methods. registry ↩

[3] Rick Durrett, Probability: Theory and Examples, 5th ed., Cambridge University Press, 2019. registry ↩

[4] Olav Kallenberg, Foundations of Modern Probability, 3rd ed., Springer, 2021. registry ↩