Bayesian model reduction¶

A method deriving evidence and posterior parameters for models with altered priors from a previously fitted full Bayesian model.

Version: v2 · 2026-09-19 · History
Domain-specific #: 3424
Domain group: Formal Sciences
Origin domain: Experimental Design & Statistics
Subdomain: Bayesian Statistics → Experimental Design & Statistics

Core Idea¶

Analytic shortcuts usually require specified distributional families such as Gaussian priors and posteriors; reduced models typically constrain or switch off parameters without refitting the likelihood.^[1] A full model is inverted once, a reduced prior encodes a nested hypothesis and a prior-posterior ratio identity updates the full evidence and posterior to score the reduced model efficiently. 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.

The load-bearing residual is not the broad topic of bayesian statistics. It is the domain-specific identity fixed by the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit. That residual remains recognizable when examples, notation, scale, or implementation change, but it disappears if the carrier is mistyped, the condition that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit fails, a neighboring object is substituted, or notation and topical resemblance replace the constitutive test. This gives the entry an operational identity rather than merely a historical label.

A useful analysis keeps three layers separate. The constitutive layer says what must be true: the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit. The evidential layer asks what observation or proof warrants the claim: type the carrier, state every parameter and convention in the definition, test that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases. The use layer asks what reasoning becomes available once the identity is established: recognizing and comparing instances of Bayesian model reduction, deriving its domain-specific consequences, selecting valid models or methods, and preventing transfer beyond its assumptions. Conflating the layers is the most common source of scope inflation.

How would you explain it like I'm…

Switch-Off Shortcut

Suppose you build a big model with lots of knobs and do the hard work of fitting it once. Now you want to know if a simpler version, with some knobs switched off, would do just as well. Bayesian model reduction is a shortcut that tells you the answer using what you already worked out, without doing all the hard work again.

Scoring Simpler Models Without Refitting

In Bayesian statistics, fitting a model to data can take a lot of computing, and you often want to compare many versions of a model. Bayesian model reduction lets you fit only the biggest, full model once. Then, to test a simpler version where some parts are turned off or limited, you just change your starting beliefs, called priors, for those parts. A math shortcut uses the full model's results to quickly score the simpler version and update its answers, without refitting to the data. The shortcut usually works exactly only when the beliefs follow certain standard shapes, like bell curves.

Scoring Nested Models Without Refitting

Bayesian model reduction is a shortcut for comparing a full Bayesian model with simpler, nested versions of it. You first fit, or 'invert', the full model once, getting its posterior and its evidence (how well it explains the data overall). A reduced model is described by changing only the prior, for example making certain parameters fixed at zero, which switches them off; the likelihood stays the same. An identity based on the ratio between the prior and the posterior then lets you compute the reduced model's evidence and posterior directly from the full model's results, without refitting to the data. The analytic formulas usually rely on specific distribution families, such as Gaussian priors and posteriors. This makes it cheap to score many nested hypotheses, as long as those assumptions are stated and checked against direct refitting where possible.

Bayesian model reduction is a method for scoring and updating nested reduced models from a single inversion of a full model. The full model's likelihood, prior, posterior and evidence are computed once. A reduced model is specified purely by a reduced prior that constrains or switches off parameters, so it shares the full model's likelihood. Because the likelihood is common, the posterior of the reduced model is proportional to the full posterior multiplied by the ratio of reduced to full priors, and the reduced evidence equals the full evidence times the full-posterior expectation of that prior ratio. Under specified distributional families, typically Gaussian priors and posteriors, these integrals are analytic, giving closed-form expressions for the evidence difference and the reduced posterior parameters. This allows large model spaces to be searched without refitting each candidate to the data. A careful application states the full model's quantities, the reduced prior and nested-parameter mapping, the distributional assumptions, the analytic evidence formula and parameter update, numerical stability, and how results compare with direct refitting.

Structural Signature¶

Carrier: the typed bayesian statistics carrier, including objects, relations, parameters, conventions, evidence, boundaries, and comparison targets
Inputs or antecedent state: the exact bayesian statistics carrier, defining parameters and conventions, boundary conditions, source evidence, comparison cases, and any measurement or proof assumptions needed to evaluate Bayesian model reduction
Constitutive operation: A full model is inverted once, a reduced prior encodes a nested hypothesis and a prior-posterior ratio identity updates the full evidence and posterior to score the reduced model efficiently.
Invariant: the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit
Recognition test: type the carrier, state every parameter and convention in the definition, test that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases
Output or consequence: recognizing and comparing instances of Bayesian model reduction, deriving its domain-specific consequences, selecting valid models or methods, and preventing transfer beyond its assumptions
Failure boundary: the carrier is mistyped, the condition that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit fails, a neighboring object is substituted, or notation and topical resemblance replace the constitutive test

What It Is Not¶

It is not the whole field of bayesian statistics. The field contains many questions and methods that do not instantiate Bayesian model reduction.
It is not its most familiar example. A canonical instance directly demonstrates that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit. exhibits the structure, but the example is evidence for the abstraction rather than its definition.
It is not the neighboring catalog concept Bayesian model selection. Bayesian model selection compares evidences across independently specified or fitted models; model reduction reuses one full-model posterior to evaluate a family of prior-restricted models.
It is not a claim that every boundary case has one uncontested classification. a generalized or degenerate case may change existence, uniqueness, measurement, or naming conventions, so the exact definition of Bayesian model reduction must control the decision
It is not an unrestricted metaphor for any process that seems similar. Outside bayesian statistics, the vocabulary and validity conditions do not transfer literally.

Scope of Application¶

Bayesian model reduction belongs to bayesian statistics and is useful where the analyst can specify the typed bayesian statistics carrier, including objects, relations, parameters, conventions, evidence, boundaries, and comparison targets, then evaluate the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit. The scope is broad within that domain but bounded by the need for the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit. The entry records a descriptive analytical identity; practical use requires the governing domain's evidence, standards, and safety obligations.^[2]

Definition and recognition. Determine whether a proposed instance satisfies the constitutive conditions rather than merely sharing terminology.
Construction or evolution. Track how the exact bayesian statistics carrier, defining parameters and conventions, boundary conditions, source evidence, comparison cases, and any measurement or proof assumptions needed to evaluate Bayesian model reduction are converted, constrained, or organized by A full model is inverted once, a reduced prior encodes a nested hypothesis and a prior-posterior ratio identity updates the full evidence and posterior to score the reduced model efficiently..
Comparison. Compare instances using carrier, parameters, convention, domain, scale, boundary conditions, evidence, exact versus approximate form, and limiting behavior, without treating convenience measures as the definition.
Boundary analysis. Diagnose cases where a generalized or degenerate case may change existence, uniqueness, measurement, or naming conventions, so the exact definition of Bayesian model reduction must control the decision and state which convention or theorem controls the decision.
Downstream reasoning. Use the established identity to support recognizing and comparing instances of Bayesian model reduction, deriving its domain-specific consequences, selecting valid models or methods, and preventing transfer beyond its assumptions while preserving the assumptions under which the inference is valid.

Clarity¶

The abstraction clarifies a crowded vocabulary by making the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting 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. A bare label is insufficient because the name Bayesian model reduction can be used for a formal identity, an implementation, or a neighboring result unless carrier and convention are stated. The disciplined statement is: given the exact bayesian statistics carrier, defining parameters and conventions, boundary conditions, source evidence, comparison cases, and any measurement or proof assumptions needed to evaluate Bayesian model reduction, the structure counts as Bayesian model reduction exactly when the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit.

This format also separates identity from measurement. Empirical, computational, or documentary proxies support recognition only under declared validity and uncertainty assumptions; formal cases require proof rather than measurement. Measurements can be noisy, implementations can approximate, and proofs can use equivalent characterizations; none of those facts licenses changing the object being measured. When reports disagree, first check scope and convention, then data or proof, and only then interpret the disagreement as substantive.

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 Bayesian model reduction. Bayesian model reduction 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.

The compression has a price. A single label can hide canonical, generalized, restricted, approximate, computational, empirical, and historically variant formulations of Bayesian model reduction. Good use therefore carries a small declaration of assumptions alongside the name. The abstraction manages complexity when it reduces the state space of the question while keeping the failure boundary visible; it mismanages complexity when the label substitutes for that boundary analysis.

Abstract Reasoning¶

Identify the carrier. State what the elements, states, objects, or observations are: the typed bayesian statistics carrier, including objects, relations, parameters, conventions, evidence, boundaries, and comparison targets. Reject examples whose alleged carrier belongs to a different problem.
Lock the constitutive rule. Express the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit independently of one notation or implementation. This step prevents the canonical example from becoming the definition.
Derive consequences. From the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit, infer recognizing and comparing instances of Bayesian model reduction, deriving its domain-specific consequences, selecting valid models or methods, and preventing transfer beyond its assumptions. Record each assumption used so that a later change of setting does not silently preserve an invalid conclusion.
Test adversarial cases. Examine a generalized or degenerate case may change existence, uniqueness, measurement, or naming conventions, so the exact definition of Bayesian model reduction must control the decision and an object that resembles Bayesian model reduction in purpose or vocabulary but does not satisfy its invariant is outside the class. A robust identity explains why the first is convention-sensitive and why the second is outside the class.
Compare and refine. Use carrier, parameters, convention, domain, scale, boundary conditions, evidence, exact versus approximate form, and limiting behavior to compare legitimate instances, and refine the model when discrepancies reflect hidden variation rather than failure of the abstraction itself.

Knowledge Transfer¶

Knowledge transfers strongly among subfields of bayesian statistics because they reuse the typed bayesian statistics carrier, including objects, relations, parameters, conventions, evidence, boundaries, and comparison targets, A full model is inverted once, a reduced prior encodes a nested hypothesis and a prior-posterior ratio identity updates the full evidence and posterior to score the reduced model efficiently., and type the carrier, state every parameter and convention in the definition, test that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases. A theorem, diagnostic, or modeling warning can travel when those roles remain literal. For example, the distinction between constitutive identity and a convenient observable transfers from A canonical instance directly demonstrates that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit. to An applied instance preserves the invariant under changed notation, scale, dataset, jurisdiction, or implementation..^[3]

Transfer outside the home domain is weaker. The skeletal pattern—type the carrier, apply the defining mechanism of Bayesian model reduction, preserve its invariant, and derive only consequences licensed by the stated boundary—may suggest an analogy, but the domain-specific mechanisms, admissible evidence, and consequences do not come along automatically. The safe transfer procedure maps each role explicitly, checks the invariant again, and refuses the name when only a superficial resemblance remains.

Examples¶

Canonical¶

A canonical instance directly demonstrates that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit. The example exposes the carrier and directly tests that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit; changing incidental notation preserves the identity, while removing that condition destroys it. This example is canonical because every role can be inspected: the carrier is the typed bayesian statistics carrier, including objects, relations, parameters, conventions, evidence, boundaries, and comparison targets; the operative rule is A full model is inverted once, a reduced prior encodes a nested hypothesis and a prior-posterior ratio identity updates the full evidence and posterior to score the reduced model efficiently.; the invariant is the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit; and the result supports recognizing and comparing instances of Bayesian model reduction, deriving its domain-specific consequences, selecting valid models or methods, and preventing transfer beyond its assumptions.^[1] Changing incidental notation or scale leaves the structure intact, while removing the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit destroys the classification.

Mapped back: the typed bayesian statistics carrier, including objects, relations, parameters, conventions, evidence, boundaries, and comparison targets → A full model is inverted once, a reduced prior encodes a nested hypothesis and a prior-posterior ratio identity updates the full evidence and posterior to score the reduced model efficiently. → the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit → recognizing and comparing instances of Bayesian model reduction, deriving its domain-specific consequences, selecting valid models or methods, and preventing transfer beyond its assumptions

Applied / In Practice¶

An applied instance preserves the invariant under changed notation, scale, dataset, jurisdiction, or implementation. The applied case qualifies only because the same invariant and boundary test remain literal under changed parameters or implementation. The applied case is not licensed merely by vocabulary. It qualifies because the same recognition test—type the carrier, state every parameter and convention in the definition, test that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit, compare the nearest accepted identity, and report counterexamples, uncertainty, and limiting cases—can be run and because the same failure boundary—the carrier is mistyped, the condition that the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting are explicit fails, a neighboring object is substituted, or notation and topical resemblance replace the constitutive test—remains meaningful.^[2] The case also shows why practical outputs should report assumptions, resolution, and uncertainty instead of a naked label.

Mapped back: declared instance → recognition test → boundary check → qualified use

Structural Tensions¶

T1: Axiomatic identity vs. operational recognition. The defining conditions may be exact while empirical or computational recognition is approximate. Neither pole can be removed without changing the analytical task. Diagnostic: Can the reviewer state both the exact condition and the evidence used to infer it?
T2: Local roles vs. global consequence. The mechanism is enacted through local relations, but the abstraction is usually valued for a global classification or prediction. Neither pole can be removed without changing the analytical task. Diagnostic: Does the claimed global result actually follow from the declared local conditions?
T3: Ideal form vs. finite representation. Theory states a clean invariant while data structures, measurements, or proofs expose only finite representations. Neither pole can be removed without changing the analytical task. Diagnostic: Would increasing resolution converge toward the same classification?
T4: Canonical convention vs. legitimate variants. A standard formulation supports communication, while variants may preserve the same core under changed assumptions. Neither pole can be removed without changing the analytical task. Diagnostic: Which role is invariant across variants, and which convention-specific conclusion changes?
T5: Compression vs. hidden assumptions. The name compresses a complex argument but can conceal prerequisites. Neither pole can be removed without changing the analytical task. Diagnostic: Can each downstream inference be traced to an explicit assumption?
T6: Autonomous residual vs. reduction to catalog neighbors. The candidate uses broader structures but adds an identity-bearing residual. Neither pole can be removed without changing the analytical task. Diagnostic: After subtracting the proposed parent and named neighbors, does the constitutive residual still support independent diagnostics?

Structural–Framed Character¶

The entry is structurally mixed but domain-framed. Its portable skeleton is type the carrier, apply the defining mechanism of Bayesian model reduction, preserve its invariant, and derive only consequences licensed by the stated boundary. Its identity-bearing terms—Bayesian model reduction, carrier, parameter, invariant, boundary, evidence, model, transformation, and application—derive their meaning from bayesian statistics and cannot be replaced by generic systems language without losing the tests that distinguish valid from invalid instances.

This mixed character explains why the abstraction is reusable inside the domain yet does not meet the Prime bar. The structure organizes reasoning, but its claims still depend on domain-specific objects, evidence, and intervention semantics.

Structural Core vs. Domain Accent¶

The structural core consists of a carrier, A full model is inverted once, a reduced prior encodes a nested hypothesis and a prior-posterior ratio identity updates the full evidence and posterior to score the reduced model efficiently., a recognition invariant, and a consequence. That skeleton may resemble patterns elsewhere, especially type the carrier, apply the defining mechanism of Bayesian model reduction, preserve its invariant, and derive only consequences licensed by the stated boundary. The domain accent is not decorative: Bayesian model reduction, carrier, parameter, invariant, boundary, evidence, model, transformation, and application determine what counts as an admissible carrier, a valid transition, and successful evidence.

The abstraction therefore remains domain-specific. A cross-domain reuse that preserves only words such as 'balance,' 'cut,' 'sequence,' 'loss,' or 'simulation' is metaphor. Literal transfer requires the original role structure and diagnostics, which in this case remain anchored in bayesian statistics.

This entry is a kind of Bayesian Updating.

It remains its own entry because its identity is fixed by the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting.

Relationships to Other Abstractions¶

Current abstraction Bayesian model reduction Domain-specific

Parents (1) — outgoing typed relations

Bayesian model reduction is a kind of Bayesian Updating Prime

It remains its own entry because its identity is fixed by the full likelihood prior posterior and evidence, reduced prior and nested-parameter mapping, distributional assumptions, analytic evidence-difference formula, parameter update, numerical stability and comparison with direct refitting.

Hierarchy paths (3) — routes to 2 parentless roots

Bayesian model reduction → is a kind of Bayesian Updating → is a kind of Inductive Reasoning

Show alternative paths (2)

Neighborhood in Abstraction Space¶

Bayesian model reduction sits in a crowded region of the domain-specific corpus (3^rd percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Statistical Inference & Hypothesis Testing (34 abstractions)

Nearest neighbors

Widely applicable information criterion — 0.95
Bayesian linear regression — 0.94
Marginal likelihood — 0.94
Normal-inverse-gamma distribution — 0.94
Posterior probability — 0.92

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

Not to Be Confused With¶

Bayesian model selection. Bayesian model selection compares evidences across independently specified or fitted models; model reduction reuses one full-model posterior to evaluate a family of prior-restricted models.
One canonical example. An instance demonstrates the structure but does not define the whole abstraction.
Measurement or implementation of Bayesian model reduction. A proxy or realization is evidence for the abstraction, not the abstraction itself.
Generalized Bayesian model reduction. An extension qualifies only when its changed axioms and retained invariant are stated.

References¶

[1] Karl Friston, Will Penny, 'Post hoc Bayesian model selection', NeuroImage, June 2011, doi:10.1016/j.neuroimage.2011.03.062. registry ↩a ↩b Show verification details

[2] Karl J Friston, Vladimir Litvak, Ashwini Oswal, Adeel Razi, Klaas E Stephan, Bernadette C.M van Wijk, 'Bayesian model reduction and empirical Bayes for group (DCM) studies', NeuroImage, March 2016, doi:10.1016/j.neuroimage.2015.11.015. registry ↩a ↩b

[3] M.J Rosa, K Friston, W Penny, 'Post-hoc selection of dynamic causal models', Journal of Neuroscience Methods, June 2012, doi:10.1016/j.jneumeth.2012.04.013. registry ↩ Show verification details