# Effect-scale and individual-conversion register

This register covers every row in the 14-row integrated policy/aggregate ledger and the 23-row individual-effect ledger. Four current headline families outside that 14-row ledger are separately specified in `headline_family_contracts.csv`, without fabricating independent result rows. It is an estimand and conversion contract, not a newly pooled causal analysis. Reported effect values remain on their original scale. The 14-row ledger is historical and some estimates are not current lead estimates; see each status and the current manuscript for disposition.

The policy estimands are not exchangeable: RR per 10 pp, RR per SD proxy, eligibility ITT, and differential-outcome DDD do not estimate one common vaccinated-versus-unvaccinated effect. The zero causal-pool flag is conservative pending a formally specified common estimand, validated design-specific first stages, cross-design bias model, and sampling covariance. It does not imply zero information or zero effect.

An ecological outcome can support a conditional individual-scale model when the cohort/area exposure path is observed and the full mixture term is fitted in the original design. This is possible for Japan and has been attempted for PCV7; it does not make those models automatically causal. Conversely, a regression coefficient divided by a nominal ±5 pp or ±SD exposure pair is generally not a valid inversion of a spline-adjusted event study.

| Translation status | Rows | Meaning |
|---|---:|---|
| design_exact_conditional_model_available | 2 | Full design-specific exposure path modeled; assumptions and uncertainty still limit causal interpretation. |
| design_weighted_forward_map_needed | 6 | Observed exposure proxy exists, but the estimator-weighted cohort exposure shift has not been validated. |
| native_only_differential_outcome | 1 | Outcome contrast is not a single autism risk ratio. |
| native_only_failed_identification | 1 | Native association remains descriptive because a pretrend diagnostic failed. |
| policy_itt_first_stage_needed | 4 | Schedule eligibility/policy contrast; receipt first stage and IV or mixture assumptions needed for individual effect. |
| reported_individual_association_native_scale | 23 | Published adjusted RR, OR, or HR retained without conversion or causal upgrade. |

The `literature_pool_flag` reproduces the legacy family-pool choice, not a causal-pooling endorsement. Shared `overlap_block` values prohibit interpreting multiple rows as independent replications. RR, OR and HR have distinct denominators/time scales; their numerical similarity is not a license for automatic pooling.

PCV7 direct mixture and signed-score audits use the same 353 state-cohort cells as the prior coefficient; they are re-analyses, not independent studies. Japan continuous-mixture estimates use the same nine source cohorts as the linear coverage gradient. Neither is added as a new independent row.

For an individual causal RR, additionally require exposure consistency, homogeneous or explicitly modeled heterogeneous effects, correct baseline-risk composition, outcome measurement stability, counterfactual trend validity, and receipt/first-stage uncertainty. For a policy IV-LATE, also require an applicable exclusion restriction, monotonicity and a nonzero validated first stage. No ecological-to-individual transformation substitutes for these assumptions.