← Every exhibit

Table Figure 1 From the research bench

The synthetic selection-bias eligibility ledger

The inclusion mechanism belongs beside the outcome. A sample label alone does not show whether the target comparison survived selection.

IDTarget and intended unitInclusion or conditioning eventSelected sampleSelection variable to inspectPossible directionDesign response
S-01All eligible trial accountsAccount completed setupSetup completers onlyBaseline capability and motivationUnknownCompare entry population; retain non-completers
S-02All invited buyersSurvey responseRespondentsInterest and response burdenUnknownTrack nonresponse; compare frame variables
S-03All routed leadsSales acceptanceAccepted leadsQualification and seller capacityCould change both waysPreserve rejected leads; model routing path
S-04All campaign targetsAd exposureExposed and holdout usersTargeting score and prior behaviorLikely selection riskRandom assignment or validated design
S-05All open opportunitiesProposal reachedProposal-stage recordsDeal maturity and seller choiceUnknownReport stage entry; do not call late-stage rate funnel-wide
S-06All retained accountsRemained observable through day 60Day-60 respondersEarly value and survivalUnknownTreat attrition as a separate outcome and sensitivity

Swipe or scroll horizontally if the table is wider than your screen.

Cite Embed

Reference & Evidence

Source: Author's synthetic eligibility ledger grounded in Lewis and Rao (2015). Targets, selection paths, possible directions, tests, and dispositions are illustrative.

Each line is a claim from the register this journal publishes against, resolved from the register at build time.

  • A Why observational methods fail here is an economics problem, in the authors' words: "The median confidence interval on return on investment is over 100 percentage points wide", so "informative advertising experiments can easily require more than 10 million person-weeks" Lewis & Rao (2015) · LR15-C3
  • B Selection bias arises when the mechanism that puts units into the observed or analyzed sample is related to variables relevant to the target question Author framework grounded in LR15-C3 · R09-OWN-C1
  • B The target population, eligibility rule, observed sample, and conditioning event are different objects and must not be collapsed into one denominator Author framework · R09-OWN-C2
  • B Conditioning on a common consequence of exposure and outcome predictors can create collider bias even when the initial variables were not associated Author causal framework · R09-OWN-C4
  • B A selection review should name the target, inclusion path, excluded units, selection variables, expected direction if known, and design or sensitivity test Author operating framework · R09-OWN-C6
  • B The six-row eligibility ledger is synthetic and contains no participant, customer, employee, or company sample Author synthetic object · R09-OWN-C7
  • B A changed eligibility rule can change the comparison set and raw rate without proving a change in the underlying outcome process Author negative boundary · R09-OWN-C9

Grades: A, verified against the printed page of the primary source · B, primary source, text layer only · C, authoritative secondary · D, reported.

Related exhibits