Reporting-Pyramid Undercount¶
Core Idea¶
Reporting-Pyramid Undercount occurs when a true population of countable events must pass through several ordered detection, classification, escalation, and recording stages before it appears in an official dataset. Each stage admits only a fraction of the events that reached it, so the final capture fraction is the product of the layerwise fractions. Because those fractions usually vary with severity, visibility, access, incentives, and administrative practice, the recorded set is not merely smaller than the true set; it is compositionally biased.
The abstraction also explains a moving-measurement problem. If a layer becomes easier to pass—because awareness rises, reporting rules change, or capacity expands—the recorded count can increase even when true incidence is unchanged. A falling count can likewise reflect reporting fatigue rather than improvement. The measured series is therefore jointly produced by the underlying event process and the reporting pipeline.
How would you explain it like I'm…
The Leaky Counting List
Lost at Every Step
The Leaky Reporting Pipeline
Broad Use¶
- Public health: infections pass through symptom, care-seeking, testing, confirmation, and notification layers.
- Pharmacovigilance: adverse reactions must be recognized, attributed, reported, accepted, and entered into a safety database.
- Occupational safety: injuries pass through worker disclosure, supervisor recording, reportability classification, and registry submission.
- Crime and abuse statistics: experienced incidents must be recognized, disclosed, accepted, classified, and recorded.
- Cybersecurity and operations: events must be detected, triaged, escalated, declared, and logged.
- Public administration: eligible cases must encounter, navigate, qualify for, and complete an administrative reporting channel.
Clarity¶
The prime separates four questions that a raw count fuses: how many events occurred, how many crossed each gate, which kinds were preferentially lost, and whether the gates changed over time. It replaces “the data undercount reality” with a layer-by-layer diagnosis.
Manages Complexity¶
A complicated reporting system becomes a sequence of named gates with capture fractions. The product of those fractions yields the aggregate capture rate; the reciprocal is an undercount multiplier when the required assumptions and external evidence are available. The decomposition also directs intervention toward the binding layer rather than toward reporting in general.
Abstract Reasoning¶
If \(N\) true events pass through \(k\) layers with capture fractions \(p_1,\ldots,p_k\), then the expected recorded count is \(R=N\prod_i p_i\). When the \(p_i\) vary across event types or people, the recorded composition is biased. When they vary over time, changes in \(R\) cannot be attributed to changes in \(N\) without separately estimating the pipeline.
Knowledge Transfer¶
The mechanism transfers whenever an in-principle countable event must cross sequential inclusion gates before entering a decision dataset. The layer labels may change completely, but the ordered filtering, multiplicative loss, composition bias, and moving-filter confound remain recognizable.
Example¶
A safety program observes a 30% rise in recorded incidents after introducing anonymous mobile reporting. The increase need not mean that the workplace became less safe. The new channel may have raised the disclosure-layer capture fraction while true incidence stayed flat. The correct analysis estimates both the event process and the changed reporting process.
Relationships to Other Abstractions¶
Current abstraction Reporting-Pyramid Undercount Prime
Parents (1) — outgoing typed relations
-
Reporting-Pyramid Undercount is a kind of Selection Bias Prime
Reporting-Pyramid Undercount is Selection Bias specialized to an ordered sequence of non-random inclusion gates whose capture fractions multiply.
Children (1) — abstractions linked here by typed relations
-
Outbreak Underascertainment Domain-specific is a kind of Reporting-Pyramid Undercount
Outbreak Underascertainment is Reporting-Pyramid Undercount specialized to infectious-disease surveillance and its symptom, care-seeking, testing, confirmation, and notification layers.
Hierarchy paths (5) — routes to 5 parentless roots
- Reporting-Pyramid Undercount → is a kind of Selection Bias → is a kind of Bias
- Reporting-Pyramid Undercount → is a kind of Selection Bias → presupposes Statistical Inference → is a kind of Inductive Reasoning
- Reporting-Pyramid Undercount → is a kind of Selection Bias → presupposes Statistical Inference → presupposes Uncertainty
- Reporting-Pyramid Undercount → is a kind of Selection Bias → is a kind of Vantage-Induced Omission → presupposes Viewpoint
- Reporting-Pyramid Undercount → is a kind of Selection Bias → presupposes Statistical Inference → presupposes Probability → is a kind of Measure → presupposes Set and Membership
Distinction from Neighbors¶
Selection Bias is the broader genus; Reporting-Pyramid Undercount fixes its mechanism to ordered gates and adds a multiplicative decomposition. Missing Data describes absence in a dataset without requiring a sequential reporting process. Attrition is loss from an already enrolled or observed population, while reporting-pyramid losses can occur before any record exists. Reporting delay concerns events that will eventually arrive; undercount concerns events permanently filtered out. Measurement Error can distort event values without changing which events are counted.