Composition-vs-Transformation Dashboard¶
Decision dashboard — instantiates Selection–Transmission Change Attribution
Displays how much of an aggregate shift is composition versus within-unit transformation and routes the decision to the matching intervention lever.
A decomposition that no decision-maker can read is inert. Composition-vs-Transformation Dashboard is the presentation-and-routing layer: it takes the selection and transmission terms an accounting mechanism already computed, shows how much of the aggregate shift is composition versus within-unit transformation — broken out by segment — and then fires a documented rule that routes the decision to the matching lever. Its defining move is that it computes no terms of its own; it consumes them and converts them into an action assignment. Selection dominance points at composition, selection pressure, and retention levers; transmission dominance points at training, process, and repair levers. The dashboard is where "the arithmetic says selection" becomes "so change the referral policy, not the clinical program."
Example¶
A national hospital network's risk-adjusted 30-day mortality rate improved over two years, and leadership wants to know what to reinforce. The dashboard ingests the split from the decomposition and renders it as a stacked contribution bar: 70% of the improvement is composition — low-mortality specialty centers grew their share of admissions — and only 30% is transformation, hospitals actually improving care. A stratified view slices the same split by region, revealing the composition effect concentrated in two metros where referral patterns shifted. The routing rule then fires: because selection dominates and rests on referral flows that could reverse, the improvement is fragile, and the recommended lever is admissions and referral policy — not another network-wide clinical-quality campaign that the numbers say would barely move the aggregate.
How it works¶
Ingest the selection term, transmission term, and residual from the accounting mechanism; render them as a contribution view (a stacked bar or waterfall) with a stratified breakdown so the split can be read per segment as well as in aggregate. Then apply the intervention attribution rule — a documented mapping from "dominant term plus sign" to a recommended lever, drawn from a fixed lever library. The dashboard stops at the recommendation. It performs no robustness sweep and states no causal claim beyond the accounting split it was handed; its whole value is turning an abstract two-number result into a legible, routed decision.
Tuning parameters¶
- Dominance threshold — how large a term's share must be before it triggers its lever. A high threshold avoids acting on marginal splits but leaves ambiguous cases unrouted.
- Stratification depth — how finely the split is sliced (region, product, cohort). Deeper slicing localizes the effect but multiplies panels and invites over-reading small strata.
- Lever library — which concrete actions map to selection dominance versus transmission dominance. A richer library gives more precise routing but demands upkeep as interventions change.
- Uncertainty display — whether the residual and any upstream sensitivity band are shown alongside the bars. Showing them tempers false confidence; hiding them makes a cleaner but more dangerous chart.
When it helps, and when it misleads¶
Its strength is legibility and forced choice: it turns a decomposition into something an executive can act on and makes the "which lever" question unavoidable rather than leaving the split to gather dust in an appendix.
Its failure mode is that a clean chart invites treating the mathematical split as causal proof and acting on what is quantified while ignoring what is not — the McNamara fallacy, where the measurable is elevated and the unmeasured is dismissed.[n1] A dominance bar showing "70% selection" says the arithmetic attributes the move there, not that the routed lever will actually shift the aggregate. The guarding discipline is to label the split as accounting rather than mechanism, to show the residual and robustness band on the same view, and to require a sensitivity check before high-stakes routing rather than acting on a single-specification bar.
How it implements the components¶
intervention_attribution_rule— the documented dominant-term-to-lever mapping it applies to convert the split into a recommended action.stratification_layer— the per-segment breakdown it renders so the split can be read within strata, not just in aggregate.
It computes neither channel itself (selection_term_definition, transmission_term_definition — those come from the Price Equation Decomposition Table and its term calculators, which it consumes) and it runs no robustness sweep (variance_and_response_capacity_register). Its nearest twin is the Selection–Transmission Sensitivity Analysis: the dashboard communicates a single point-estimate split and routes it to a lever, while the sensitivity analysis stress-tests whether that split survives alternative choices — the separating component is the intervention-routing rule it owns versus variance_and_response_capacity_register.
Related¶
- Instantiates: Selection–Transmission Change Attribution — it is the mechanism that makes the split legible to decision-makers and routes action.
- Consumes: Price Equation Decomposition Table supplies the selection and transmission terms it displays.
- Sibling mechanisms: Price Equation Decomposition Table · Covariance Selection-Term Calculation · Within-Unit Change Assay · Lineage or Panel Correspondence Matrix · Entry/Exit Normalization Protocol · Decomposition Residual Reconciliation Workflow · Selection–Transmission Sensitivity Analysis
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: The mechanism renders supplied decomposition terms and applies a documented mapping from dominant term and sign to a recommended intervention lever, turning an accounting result into decision-support analysis.
Nearest alternative: Decision, Gate & Allocation — The readout routes a recommendation, but it explicitly stops before acting or allocating and leaves the final lever choice to a separate decision-maker.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Data Science & Analytics
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Analytic-dashboard practice supplies the presentation and rule-based routing of decomposed aggregate change into decision levers.
Related originating lineages:
- Biology & Ecology — The Price equation supplies the source's named selection and transmission terms and their composition-versus-within-lineage interpretation.
- Economics & Finance — Shift-share and opportunity-index decomposition independently operationalize composition-versus-within-group effects for policy decisions.
- Statistics & Experimental Design — Statistical decomposition supplies contribution accounting, residual reconciliation, and stratified uncertainty handling.
Review resolution: The source explicitly computes nothing; its defining operation is a dashboard that renders precomputed components and routes them to action. The World Bank's Equity Lab is an authoritative example of a dashboard presenting a change decomposition into composition and coverage effects, while Price's primary paper anchors the selection/transmission terms consumed here. Data science is therefore primary for the presentation-and-routing method, with biology, statistics, and economics supplying genuine decomposition lineages.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- World Bank LAC Equity Lab: Human Opportunity Index and change decomposition
- George R. Price, Selection and Covariance
Notes¶
[n1] The McNamara fallacy is the error of making decisions solely on quantifiable metrics while disregarding what cannot easily be measured — named for the Vietnam-era reliance on body counts. A dashboard that routes action from a tidy split is especially exposed to it, which is why the unquantified and the residual must stay on the view. ↩