The popdgp methods paper: formalization + metric-blind-spot power study + multi-view

Task

The popdgp methods paper: "Evaluating synthetic populations against the surveys that observe them." Venue: Survey Methodology or Journal of Official Statistics (both diamond OA); submits after imputation-paper (which it cites as the first application).

Contributions

1. Formalization: latent population, survey views (variable subset ∘ design ∘ measurement), per-view scoring in the view's own idiom (no cross-survey consistency required), sampling floors, and the block decomposition — in particular coverage's invariance to candidate reweighting (the calibration-blind block) and what that separates.
2. Designed metric-blind-spot study: plant known failures (tail inflation ×k, mode collapse, broken coupling, wrong zero mass, measure-vs-support errors) in synthetic populations and measure which blocks detect which failures at which sample sizes and subsample caps. The imputation-paper's tail-blindness discovery (capped geometry certifying a 2× q99 error), generalized from anecdote to power study.
3. Multi-view instantiation: CPS + SCF (+ SIPP when data lands) simultaneously.
4. Formal-ancestry positioning (data integration/superpopulation; block-missingness fusion; proper scoring rules; PRDC) — no unified operational benchmark exists in the population-synthesis literature.

Dependencies: the extraction issue; SIPP source for the third view (optional for v1).

🤖 Generated with Claude Code

Start by reading the extraction issue, then inspect the repository entry points for the evaluation harness and its existing survey inputs. Define the formal view and scoring model, implement the metric-blind-spot power study, and instantiate it for CPS and SCF; SIPP is optional for v1. Done means the methods paper's formalization, study results, and multi-view evaluation are reproducible.