Cox process¶

A point process that is conditionally Poisson given a random intensity measure, thereby representing clustered or environment-driven event rates.

Version: v2 · 2026-09-19 · History
Domain-specific #: 3955
Domain group: Formal Sciences
Origin domain: Mathematics
Subdomain: Point Processes → Mathematics

Core Idea¶

Also called a doubly stochastic Poisson process, it separates randomness of the directing intensity from conditional Poisson event noise; log-Gaussian and shot-noise Cox processes specify different directing measures. A stochastic field or random measure is drawn first, then events are generated as an inhomogeneous Poisson process conditional on that realized intensity. The abstraction is therefore identified by a declared carrier, a transformation or constraint over that carrier, and an invariant that tells an analyst whether the named structure is genuinely present.

How would you explain it like I'm…

Two-Step Random Rain

Imagine the weather first decides how rainy each part of the sky will be; some parts get lots of clouds, some only a few. Then raindrops fall at random, more where it's cloudier. A Cox process is like that: first a random 'how busy' map, then random events following it.

Random Map, Then Random Dots

A Cox process is a way to describe random events scattered in space or time, like where trees grow or when calls arrive. It has two steps of randomness. First, a random 'busyness map' is chosen, saying how likely events are at each place. Then, given that map, the events are scattered at random according to it, the way a simple random-scatter model would do with that map. This lets you tell apart two sources of randomness: how busy each area happens to be, and where exactly the events land.

Doubly Stochastic Poisson Process

A Cox process, also called a doubly stochastic Poisson process, is a model for random points in space or time. First a random intensity, a random function or measure saying how dense events should be at each location, is drawn. Then, given that realized intensity, the points form an inhomogeneous Poisson process, meaning they are scattered independently with density following that intensity. This separates two kinds of randomness: variation in the underlying intensity, and the Poisson 'noise' of the events themselves. Because the intensity varies randomly, points tend to cluster more than in a plain Poisson process. Examples include log-Gaussian Cox processes, where the log of the intensity is a Gaussian random field, and shot-noise Cox processes, which use a different kind of random intensity.

A Cox process, or doubly stochastic Poisson process, is a point process defined hierarchically on some underlying space. A random directing measure Λ, often given by a random intensity field, is drawn from a specified law; conditional on its realization, the points form a Poisson process with mean measure Λ, independent across disjoint regions given Λ. The construction separates variability of the directing intensity from the conditional Poisson noise of event placement, so the unconditional process is overdispersed and typically clustered relative to a Poisson process with the same mean. Specifying one requires the underlying space, the directing measure and its law, local finiteness, and the resulting first- and second-order quantities such as the intensity and pair-correlation function. Choices of directing measure define families: a log-Gaussian Cox process takes the intensity to be the exponential of a Gaussian field, while a shot-noise Cox process builds it from kernels centered on random points. Stationarity, when claimed, must hold for the directing measure, and inference or simulation methods depend on these assumptions.

Scope of Application¶

Cox process belongs to point processes and is useful where the analyst can specify the typed point processes carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, then evaluate the underlying space, random directing measure and its law, conditional Poisson definition, local finiteness, intensity and pair-correlation quantities, stationarity if claimed and inference or simulation assumptions are explicit. The scope is broad within that domain but bounded by the need for the underlying space, random directing measure and its law, conditional Poisson definition, local finiteness, intensity and pair-correlation quantities, stationarity if claimed and inference or simulation assumptions are explicit. The entry records a descriptive analytical identity; practical use requires the governing domain's evidence, standards, and safety obligations.

Clarity¶

The abstraction clarifies a crowded vocabulary by making the underlying space, random directing measure and its law, conditional Poisson definition, local finiteness, intensity and pair-correlation quantities, stationarity if claimed and inference or simulation assumptions are explicit the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test.

Manages Complexity¶

Without the abstraction, an analyst must reason directly over many local details: the carrier roles, admissibility assumptions, competing conventions, derived invariants, boundary cases, and proof or validation obligations specific to Cox process. Cox process compresses them into the roles in the structural signature. That compression permits comparison across instances without erasing the variables that determine validity. It also exposes which details may be varied safely and which are constitutive.

Abstract Reasoning¶

Identify the carrier. State what the elements, states, objects, or observations are: the typed point processes carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets. Reject examples whose alleged carrier belongs to a different problem.
Lock the constitutive rule. Express the underlying space, random directing measure and its law, conditional Poisson definition, local finiteness, intensity and pair-correlation quantities, stationarity if claimed and inference or simulation assumptions are explicit independently of one notation or implementation.

Knowledge Transfer¶

Knowledge transfers strongly among subfields of point processes because they reuse the typed point processes carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, A stochastic field or random measure is drawn first, then events are generated as an inhomogeneous Poisson process conditional on that realized intensity., and type the carrier, state every parameter and convention in the definition, test that the underlying space, random directing measure and its law, conditional Poisson definition, local finiteness, intensity and pair-correlation quantities, stationarity if claimed and inference or simulation assumptions are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions¶

Current abstraction Cox process Domain-specific

Parents (1) — outgoing typed relations

Cox process is a kind of Stochastic Process Prime

It remains its own entry because its identity is fixed by the underlying space, random directing measure and its law, conditional Poisson definition, local finiteness, intensity and pair-correlation quantities, stationarity if claimed and inference or simulation assumptions.

Hierarchy path (1) — routes to 1 parentless root

Cox process → is a kind of Stochastic Process

Neighborhood in Abstraction Space¶

Cox process sits in a crowded region of the domain-specific corpus (22^nd percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Stochastic Processes & Martingale Theory (19 abstractions)

Nearest neighbors

Stationary process — 0.91
Nearest neighbour distribution — 0.90
Continuous-time stochastic process — 0.90
Excursion probability — 0.90
Stochastic drift — 0.89

Computed from structural-signature embeddings · 2026-10-25