Parent plan: #16
Goal: produce the first release-track mp-300k candidate for replacing eCPS as the routine national PolicyEngine dataset.
Important construction note: the first mp-300k candidate is a directly sampled small build, without L0 culling from a larger parent universe. That is acceptable for the first replacement test, but it should be recorded as a limitation. The long-run mp-300k path should improve by selecting/culling from a larger parent universe such as mp-3m or mp-30m.
Deliverables:
- Build small ASEC + ACS100k Microplex candidate with PUF, SIPP/SCF/Arch additions, Forbes fixed spine, and capital-gains lots enabled.
- Pin the comparison baseline: eCPS H5 path/SHA256, policyengine-us-data commit, policyengine-us version, target DB path/SHA256.
- Score against latest broad targets on common kept targets.
- Write top target deltas and protected-family deltas.
- Save config, score artifacts, target deltas, record counts, nonzero weights, ESS, H5 size, and source/target provenance in the standard artifact bundle.
Exit criterion: dashboard shows mp-300k versus pinned eCPS with enough evidence to decide release-track versus another modeling iteration.
Parent plan: #16
Goal: produce the first release-track
mp-300kcandidate for replacing eCPS as the routine national PolicyEngine dataset.Important construction note: the first
mp-300kcandidate is a directly sampled small build, without L0 culling from a larger parent universe. That is acceptable for the first replacement test, but it should be recorded as a limitation. The long-runmp-300kpath should improve by selecting/culling from a larger parent universe such asmp-3mormp-30m.Deliverables:
Exit criterion: dashboard shows
mp-300kversus pinned eCPS with enough evidence to decide release-track versus another modeling iteration.