The Observation Theory EncyclopediaFrom TSKAboutBy kindBy chapterBy Lean fileLedgerProvenance

Simpson's paradox concept

DefinitionA reversal of a comparison when groups are combined, which the book reads as an aggregation quotient. Chapter 5. Also Simpson.
ExampleA treatment better in both the young and the old can be worse overall when the two groups are mixed unequally.
BookData Mining as Observation, draft 0.2, commit f3914f0; entry id simpsons-paradox, kind concept.
Statusno ledger row names this entry. Corrections: none recorded.
Defining equationnone
Assumptions and scope
  • A pooled rate is the size-weighted mean of the group rates, so pooling is a quotient that declares the group labels irrelevant, and it can reverse a comparison that holds in every group when the groups have different sizes under the two arms.
  • With equal group sizes under both arms the pooled comparison agrees with a comparison that holds in every group. The book’s answer is the min-over-strata verdict, which never pools.
Prior artnone recorded
Evidencelean/DataMiningAsObservation/Simpson.lean
Reviewednot yet reviewed; generated 2026-09-10 from records at the commits on the provenance page.
00.51stratumscoreg1g2g3barmean passes, the minimum fails
Every group shows one sign and the aggregate shows the other.

Equation

none

Conditions

Conditions are curated in entries.toml rather than read from a record.

Ledger

none

First stated

TSK chapter 5 on objective measures, as chapter 5 section 5.5 of Data Mining as Observation reads it, an aggregation quotient.

Measurements

Where the book states it Numbers, as the book’s sources table records them Source
chapter 5 section 5.1 to 5.3, 5.5 support, confidence, lift, the Apriori principle, FP-growth, closed and maximal itemsets, objective measures, Simpson’s paradox, cross-support TSK 2e chapter 5

Failures and corrections

none

Invariance envelope

none declared

Machine checked

lean/DataMiningAsObservation/Simpson.lean, theorems pooled_eq_weighted, reversal, no_reversal_of_equal_sizes, at observation-data-mining f3914f0; what the check covers is stated in the book’s appendix C.

Used in

Data Mining as Observation primer S, chapters 5.

Related

quotient; min-over-strata; abstention; deployment mismatch.

See also

Book equations stated beside the entry’s terms, not defining it: 10.7.

Status

Generated 2026-09-10 by encyclopedia/generate.py; book at observation-data-mining f3914f0; the commit of every record is listed in the encyclopedia’s provenance.

← silhouettesingular value →