Consistency (statistics)¶

An asymptotic property in which a statistical estimator, test, interval, or other procedure approaches the correct target or decision as sample information grows.

Version: v2 · 2026-09-19 · History
Domain-specific #: 3854
Domain group: Formal Sciences
Origin domain: Experimental Design & Statistics
Subdomain: Statistical Asymptotics → Experimental Design & Statistics

Core Idea¶

Consistency distinguishes long-run correctness from finite-sample bias, efficiency, calibration, or robustness and must name the convergence mode and growth regime. The procedure is indexed by sample size or information level; under the assumed data-generating model, its error probability or distance from the target tends to zero along the declared asymptotic sequence. 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…

More Scoops, Closer Guess

Imagine guessing how many red candies are in a giant jar by scooping some out and counting. A good way of guessing is one where, the more candies you scoop, the less chance there is that your guess is far from the real answer. If you could keep scooping forever, the chance of being far off would shrink all the way to zero. That is what makes a guessing method consistent.

Getting It Right in the Long Run

In statistics, an estimator is a recipe for guessing an unknown number from data, like guessing the average height of all kids from a sample. The recipe is consistent if, as you collect more and more data, its guess gets closer and closer to the true value, with the chance of being far off shrinking toward zero. This is a long-run promise. A consistent recipe can still be a bit off with small samples, and it only works if the assumptions about how the data were made are true. Statisticians also have to say exactly what "getting closer" means, since there's more than one way to measure it.

Asymptotic Convergence to the Truth

A statistical procedure, such as an estimator or a test, is consistent if, as the sample size or amount of information grows, its error goes to zero under the assumed model for how the data are generated. For an estimator, this usually means the probability of being farther than any small distance from the true value tends to zero. Consistency is a long-run property, different from being unbiased (correct on average at a fixed sample size), efficient (low variance), calibrated, or robust to bad assumptions. A consistent estimator can be biased in small samples, and an unbiased one need not be consistent. Because there are different kinds of convergence and different ways for data to grow, a consistency claim must say which convergence mode and growth path it means, and what is assumed about the data.

Consistency (statistics) is the property that a sequence of procedures, such as estimators indexed by sample size or information level, becomes correct in the long run: under the assumed data-generating model, its error probability or distance from the target tends to zero along a declared asymptotic sequence. It concerns long-run correctness and is distinct from finite-sample properties such as bias, efficiency, calibration, and robustness; a consistent estimator can be biased at every finite n, and an unbiased one need not be consistent. Any consistency claim must name the convergence mode, for example weak consistency via convergence in probability versus strong consistency via almost-sure convergence, and the growth regime, meaning how sample size or information increases, including cases where the number of parameters grows too. It must also state the target, the procedure sequence, the data-generating assumptions, and how nuisance parameters are handled. If those are not explicit, or the error does not vanish along the specified path, the claim does not hold as stated.

Scope of Application¶

Consistency (statistics) belongs to statistical asymptotics and is useful where the analyst can specify the typed statistical asymptotics carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, then evaluate the target, procedure sequence, data-generating assumptions, sample-growth path, convergence mode, and treatment of nuisance parameters are explicit and the relevant error vanishes asymptotically. The scope is broad within that domain but bounded by the need for the target, procedure sequence, data-generating assumptions, sample-growth path, convergence mode, and treatment of nuisance parameters are explicit and the relevant error vanishes asymptotically. 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 target, procedure sequence, data-generating assumptions, sample-growth path, convergence mode, and treatment of nuisance parameters are explicit and the relevant error vanishes asymptotically the center of the account. A claim should name the carrier, the governing operation or relation, the applicable assumptions, and the recognition test. A bare label is insufficient because the name Consistency (statistics) can be used for a formal identity, an implementation, or a neighboring result unless carrier and convention are stated.

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 Consistency (statistics). Consistency (statistics) 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 statistical asymptotics 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 target, procedure sequence, data-generating assumptions, sample-growth path, convergence mode, and treatment of nuisance parameters are explicit and the relevant error vanishes asymptotically independently of one notation or implementation.

Knowledge Transfer¶

Knowledge transfers strongly among subfields of statistical asymptotics because they reuse the typed statistical asymptotics carrier, defining objects and relations, parameters, conventions, evidence, boundary cases, and comparison targets, The procedure is indexed by sample size or information level; under the assumed data-generating model, its error probability or distance from the target tends to zero along the declared asymptotic sequence., and type the carrier, state every parameter and convention in the definition, test that the target, procedure sequence, data-generating assumptions, sample-growth path, convergence mode, and treatment of nuisance parameters are explicit and the relevant error vanishes asymptotically, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases.

Relationships to Other Abstractions¶

Current abstraction Consistency (statistics) Domain-specific

Parents (1) — outgoing typed relations

Consistency (statistics) is a kind of Convergence Prime

What makes it its own entry: the domain-specific identity determined by the target, procedure sequence, data-generating assumptions, sample-growth path, convergence mode, and treatment of nuisance parameters are explicit and the relevant error vanishes asymptotically.

Hierarchy path (1) — routes to 1 parentless root

Consistency (statistics) → is a kind of Convergence

Neighborhood in Abstraction Space¶

Consistency (statistics) sits in a crowded region of the domain-specific corpus (8^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Sampling, Monte Carlo & Estimation (23 abstractions)

Nearest neighbors

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