Skip to content

Selection Transmission Change Attribution

When an aggregate mean changes, split the change into how much came from units gaining or losing weight and how much came from units changing internally.

Version
v1 · 2026-08-24 · History
Solution archetype #
942
Problem family
Scale, Hierarchy & Emergence Mismatch
Problem subfamily
Cross-Scale Attribution & Aggregation Error

When This Archetype Applies

Complete catalog groundingAt least one sufficient condition set is fully represented by existing primes or domain-specific abstractions.

A population-level weighted mean changes, but the observer cannot tell whether the shift is caused by differential weighting of units, within-unit change, or a mixture of both, so explanations and interventions target the wrong level.

Applicability expression5 distinct conditions

Aggregate change observedandUnequal unit weightsandWithin-unit changeandComposition shiftsandDriver determines response
Algebraic12345

groundedpartly groundedopen

5 conditions, all required.

5Required in every casenumbered 1–5

These hold no matter which pattern applies.

1

Aggregate change observed · grounded

An average, rate, score, trait, performance metric, risk level, or capability changes across two states or periods.

primeSelection Vs Transmission Decomposition— A change in a population's weighted mean splits exactly into a selection term (differential weighting of units) and a transmission term (units changing within themselves).

2

Unequal unit weights · grounded

Units have unequal weights, shares, survival probabilities, exposure, participation, frequency, or influence.

primeSelection Vs Transmission Decomposition— A change in a population's weighted mean splits exactly into a selection term (differential weighting of units) and a transmission term (units changing within themselves).

3

Within-unit change · grounded

Units can also change internally between states through learning, degradation, adaptation, mutation, repair, or behavioral response.

primeSelection Vs Transmission Decomposition— A change in a population's weighted mean splits exactly into a selection term (differential weighting of units) and a transmission term (units changing within themselves).

4

Composition shifts · grounded

Entry, exit, growth, shrinkage, or differential persistence changes which units dominate the aggregate.

primeSelection Vs Transmission Decomposition— A change in a population's weighted mean splits exactly into a selection term (differential weighting of units) and a transmission term (units changing within themselves).

5

Driver determines response · grounded

The needed response differs depending on whether composition or within-unit transformation is driving the change.

primeSelection Vs Transmission Decomposition— A change in a population's weighted mean splits exactly into a selection term (differential weighting of units) and a transmission term (units changing within themselves).

Other requirements and context (1)

Why these sit outside the expression

Supporting contextit may accompany or help interpret the situation, but it is not a load-bearing condition in a sufficient diagnostic set.

  • Supporting contextDecision-makers are tempted to interpret the aggregate change as a uniform population improvement or decline.

5 of 5 conditions grounded.

Read the methodologyDownload the trigger-logic data

Working definition

Selection–Transmission Change Attribution explains a change in a weighted population mean by separating two channels that are easy to confuse. The selection channel changes the aggregate because units gain, lose, enter, exit, survive, expand, shrink, or otherwise change weight. The transmission channel changes the aggregate because continuing or linked units change their own values.

This matters because the same aggregate movement can demand different interventions. If the selection term dominates, the system changed because the mix of units changed. If the transmission term dominates, the units themselves changed. If the terms have opposite signs, the aggregate may be masking one channel with the other.

When to use it

Use this archetype when an average, rate, trait, score, or capability changes over time and the units behind the aggregate have both weights and values. The pattern is especially useful when response choices differ across levels: change admission, retention, portfolio mix, survival pressure, or exposure if selection dominates; change training, repair, learning, adaptation, process design, or transmission pathways if within-unit change dominates.

Do not use it as a generic synonym for decomposition. It requires a weighted mean, a comparison across states, and a meaningful distinction between unit weights and unit values.

Core components

ComponentDescription
Population Unit and Scope Definition The population must be bounded and the units must be named. In evolutionary settings, the units may be lineages or variants. In organizations, they may be teams, products, clients, cohorts, or assets. In policy analytics, they may be regions, households, institutions, or cases. The unit rule determines what can count as selection and what can count as transmission.
Unit Value Measure The value measure is the trait, score, payoff, performance level, risk, or state being averaged. It must be comparable across the two states. If the measure changes definition halfway through, the transmission term may become a measurement artifact.
Unit Weight Measure The weight measure tells how much each unit contributes to the aggregate. Weights may represent frequency, survival, population share, revenue share, exposure, market share, case load, representation, or influence. The selection term is about changes in these weights.
Unit Correspondence Map The correspondence map links units across states. It records continuing units, new units, exited units, split units, merged units, and descendant units. Without this map, the analysis cannot reliably distinguish a unit changing internally from one unit being replaced by another.
Selection and Transmission Term Definitions The selection term accounts for aggregate change due to differential weighting. The transmission term accounts for aggregate change due to within-unit value change. Both terms must be defined in a way that recomposes to the observed aggregate change, with residuals explicitly investigated.

Common mechanisms

A Price Equation Decomposition Table is the canonical mechanism when the data support exact unit-level accounting. A Covariance Selection-Term Calculation is useful when the selection term can be expressed as covariance between unit value and relative weight change. A Lineage or Panel Correspondence Matrix helps operationalize unit continuity across states. A Composition-vs-Transformation Dashboard makes the split legible to decision-makers. A Residual Reconciliation Workflow keeps unmatched units, measurement drift, and normalization errors from being interpreted as substantive change.

Invariants to preserve

The most important invariant is recomposition: the reported terms plus declared residuals must recover the observed weighted-mean change. The second invariant is separation: unit values and unit weights must not collapse into one informal story. The third invariant is reviewability: another analyst should be able to see how units were matched, how weights were normalized, and how entry or exit was handled.

Boundary distinctions

This archetype is adjacent to aggregation-bias correction, but it is not the same thing. Aggregation-bias work asks whether an aggregate summary misleads relative to subgroup or partition structure. Selection–Transmission Change Attribution asks why an aggregate changed across states.

It is also adjacent to generic decomposition, but it is narrower and more exact. The decomposition is not arbitrary; it is organized around weighted population means, differential unit weights, and within-unit value change.

It can use panel or time-series data, but it is not equivalent to time-series cross-section analysis. Panel analysis supports many statistical questions; this archetype is specifically about attributing weighted-mean change between selection and transmission channels.

Examples

In evolutionary dynamics, a trait mean can rise because high-trait lineages leave more descendants and because descendants shift trait values relative to parents. In a company, average productivity can rise because high-performing units grow headcount, because each unit improves, or both. In education, a school’s average test score can rise because the cohort changed or because continuing students learned more. In product analytics, average engagement can rise because heavy users become a larger share, not because each segment became more engaged.

Failure modes

The most common failure is treating aggregate improvement as within-unit improvement when it is really composition change. The second is treating a mathematical split as causal proof. The third is hiding attrition or exclusion inside a favorable aggregate. The fourth is using the decomposition despite weak unit correspondence or drifting measurement scales.

Review note

The draft should be reviewed against aggregation_bias_detection_and_correction, aggregation_function_design_and_weighting, solvable_baseline_decomposition, and future coverage of variance_bounds_selection_response. The current disposition check supports a full draft because no accepted archetype or prior queue output directly covers the exact selection/transmission split.

Common Mechanisms

8 documented mechanisms across 1 implementation form.

The grouping reflects forms represented among the mechanisms currently documented for this archetype; an absent form is not necessarily an impossible implementation.

Analysis, Modeling & Optimization · 8 mechanisms

  • Composition-vs-Transformation Dashboard — Displays how much of an aggregate shift is composition versus within-unit transformation and routes the decision to the matching intervention lever.
  • Covariance Selection-Term Calculation — Isolates the selection channel by computing the covariance between a unit's value and its change in relative weight — a single statistic whose sign says whether high-value units gained share.
  • Decomposition Residual Reconciliation Workflow — Takes the leftover after selection and transmission are subtracted from the observed change and attributes it to unmatched units, scale drift, or normalization rather than substance.
  • Entry/Exit Normalization Protocol — Fixes how entrants and exiters enter the weights so that churn in the population does not masquerade as real change in the weighted mean.
  • Lineage or Panel Correspondence Matrix — Maps which units in the first state correspond to which in the second — continuing, entered, exited, split, or merged — so selection and transmission can be told apart at all.
  • Price Equation Decomposition Table — Lays out every unit's weight and value in both states as a ledger and recomposes the weighted-mean change into an exact selection term plus a transmission term.
  • Selection–Transmission Sensitivity Analysis — Re-runs the selection–transmission split under alternative windows, unit definitions, and weighting schemes to report how stable the verdict is before it drives a decision.
  • Within-Unit Change Assay — Measures the transmission channel directly by pairing each continuing unit's before and after value and averaging the within-unit change, ignoring composition entirely.

Compression statement

Selection–Transmission Change Attribution is the intervention pattern for explaining a weighted-mean shift by explicitly separating differential representation of units from within-unit transformation, then routing action to composition, selection pressure, retention, learning, repair, or measurement according to the dominant term.

Canonical formula: Δ mean ≈ selection_term + transmission_term; selection_term tracks differential unit weights, transmission_term tracks within-unit value change, with residuals audited for unmatched units and measurement drift.

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (4)

  • Aggregation: Deliberately collapsing many items into a single summary, choosing which information to discard to gain tractability.
  • Decomposition: Breaking a whole into parts that can be analyzed independently and recombined to reconstitute the whole, making complexity tractable through divide-and-conquer.
  • Natural Selection: A population of varying, heritable variants is filtered by a selection pressure so that the better-performing variants differentially reproduce or persist, shifting the population's composition over rounds — variation, selection, and retention as a substrate-neutral engine.
  • Selection Vs Transmission Decomposition: A change in a population's weighted mean splits exactly into a selection term (differential weighting of units) and a transmission term (units changing within themselves).

Also references 16 related abstractions

  • Causality: Cause-effect relationships.
  • Comparison: Place items in a shared frame along chosen dimensions to read off a relation between them.
  • Data Integrity: Accuracy and consistency preserved.
  • Diversification: Spreading exposures across positions whose failure modes are uncorrelated reduces total-outcome variance; correlation, not count, drives the benefit.
  • Effect Size: Magnitude of effect.
  • Invariance: Properties unchanged under transformation.
  • Modifiable Areal Unit Problem: Statistics computed on aggregated data change, sometimes reversing sign, when the boundaries used to aggregate are redrawn — the partition is a non-neutral analytical input.
  • Partition Dependence of Aggregates: Any statistic computed on partition-aggregated data is a function of the partition itself, not solely of the underlying data.
  • Proportionality: Match response to scale.
  • Scale: Properties change with size.

Variants

Narrower or domain-specific specializations that share this archetype's core structure. Recognized variants are established; candidate variants are provisional.

Price Equation Attribution · method family variant · recognized

A formal variant using the Price-equation style identity to split aggregate change into covariance/selection and expectation/transmission terms.

  • Distinct from parent: The parent includes any practical selection/transmission attribution; this variant emphasizes the canonical Price-equation mathematical form.
  • Use when: The population can be represented as units with weights and values across states; Differential persistence or representation and within-unit change are both plausible causes of aggregate change.
  • Typical domains: evolutionary dynamics, organizational analytics, economics
  • Common mechanisms: price equation decomposition table, covariance selection term calculation, selection transmission sensitivity analysis

Composition-vs-Capability Change Split · domain variant · recognized

A management and analytics variant that asks whether an aggregate metric changed because the mix of units changed or because each unit improved or degraded.

  • Distinct from parent: The parent is cross-domain and includes evolutionary lineages, epidemiology, economics, and learning systems.
  • Use when: A portfolio, team, customer base, fleet, or cohort changes average performance across periods; The decision response differs depending on whether the driver is mix shift or within-unit transformation.
  • Typical domains: organizational management, product analytics, public policy
  • Common mechanisms: composition vs transformation dashboard, entry exit normalization protocol, decomposition residual reconciliation workflow

Selection-Response Bounds Linkage · scale variant · candidate

A variant that connects the observed selection term to available variance and selection intensity so expected response bounds can be checked.

  • Distinct from parent: The parent attributes current aggregate change; this variant adds response-bound interpretation over repeated selection rounds.
  • Use when: The selection contribution is large enough that available variance may constrain future response; A system is being managed for adaptive response over repeated rounds rather than a single before/after attribution.
  • Typical domains: breeding, algorithmic selection, portfolio management
  • Common mechanisms: covariance selection term calculation, selection transmission sensitivity analysis

Near names: Selection Vs Transmission Decomposition, Price Equation Decomposition, Between–Within Change Decomposition, Composition–Transformation Split, Weighted Mean Change Decomposition.

Editorial Notes

Problem Classification

Classification: Scale, Hierarchy & Emergence MismatchCross-Scale Attribution & Aggregation Error

Problem kernel: aggregate change conflates reweighting with within-unit change

Rationale: Earliest causal condition: A population-level weighted mean changes, but the observer cannot tell whether the shift is caused by differential weighting of units, within-unit change, or a mixture of both, so explanations and interventions target the wrong level.

Independent corroboration: The earliest necessary condition in the frozen evidence is: A population-level weighted mean changes, but the observer cannot tell whether the shift is caused by differential weighting of units, within-unit change, or a mixture of both, so explanations and interventions target the wrong level. That is a cross scale attribution and aggregation error problem because Evidence or explanation at one level is projected onto another, hiding subgroup heterogeneity, marginal change, contingency, or part–whole causation.

Review outcome: Independent reviewer agreement; high confidence.