Completed example — Evidence Ledger

Project: PISA moving-benchmark analysis
Audit basis: Mature empirical argument

paragraph_id	principal_proposition	warrant_type	supporting_object	claim_class	required_qualification	contract_status	audit_status
P1	International rank movement can diverge from own-score movement	manuscript-generated evidence + analytic reasoning	Multi-year system means and ranks	Descriptive/interpretive	Comparison field also changes	Authorised	Pass
P2	The four UK administrations should be treated separately in this analysis	design reasoning + primary data structure	Final sample specification	Design	Aggregate UK excluded	Authorised	Pass
P3	PC1 captures approximately 97.31% of variation in the three domain means	manuscript-generated evidence	PCA output	Descriptive	Global three-domain structure	Authorised	Pass
P4	Ball Mapper adds local overlapping neighbourhood/profile representation	manuscript-generated evidence + analytic reasoning	Ball Mapper + comparator outputs	Methodological	Complement to PCA, not replacement	Authorised	Pass
P5	Broad downward movement survives landmark-order sensitivity analysis	manuscript-generated evidence	250-order robustness archive	Robustness	Exact state assignments vary	Authorised	Pass
P6	Endpoint occupancy changes do not identify direct B4→B9 same-system flow	analytic reasoning	Membership structure + revision record	Inferential boundary	Direct coupling unidentified	Authorised	Pass
P7	Rank gains can coexist with own-score declines	manuscript-generated evidence	System trajectories	Descriptive	Avoid causal interpretation	Authorised	Pass
P8	The PISA example illustrates rather than validates reason provenance	conceptual boundary	Reason-provenance manuscript	Conceptual	No empirical validation claim	Authorised	Pass

Warrant discipline

The ledger separates manuscript-generated evidence from transparent analytic reasoning.

A proposition should return to claim authorisation or analysis if its supporting object changes.