DDCM Implementation Reports
Technical progress reports from the ddcm
implementation — estimation runs, simulation results, diagnostic analyses.
These are Master’s-era reports, archived. They were generated in the ddcm
repo and published here between May and August 2026.
Current deliverables (synthetic-data methodology)
| Report | Topic | Date |
|---|---|---|
| recovery_smallnet_R30_20260620.html | Report 1 · Parameter Recovery & Identification — Exact NFXP recovers known parameters: Part I production-fidelity single-transport-scale confirmation (transport-scale ridge broken, 13.8%→0.8% bias); Part II controlled R=30 Dekker table, N-sweep (√N), profile-LL, behavioral validation, BHHH SEs | 2026-06-20 |
| experiments_exact_sa_rl_20260621.html | Report 2 · Approximation-Error Experiments — Exact vs Sampling-of-Alternatives vs RL value estimation: SA is consistent (variance only, McFadden); RL is inconsistent at finite budget (bias + variance, no correction). Variance/bias panels + per-parameter densities + bit-level oracle | 2026-06-21 |
Earlier reports
| Report | Topic | Date |
|---|---|---|
| profile_ll_cchange_v7.html | Profile LL: c_change sweep (v7 zone-agnostic, 569 persons): interior MLE at −0.615; checkpoint (−0.756) confirmed near-optimal; HH prior (−0.301) firmly rejected | 2026-05-13 |
| investigation_overview.html | Root cause investigation: zone-matching bug caused δ to be unidentified in all prior runs (v1–v5); fix implemented; re-estimation running | 2026-05-12 |
| profile_ll_delta_muhome.html | K=10 profile LL scan (δ × μ_home): confirms U_obs was δ-invariant (symptom of zone bug); interactive heatmap + slices | 2026-05-12 |
| gradient_bug_investigation_20260601.html | Gradient bug investigation — analytical gradient diagnosed against finite differences | 2026-06-01 |
| work_timing_analysis.html | K=10 NFXP: why simulated work start is ~14:51 vs observed ~8:30; interactive utility plots + simulation validation (N=934) | 2026-05-10 |
How reports are generated
HTML reports are self-contained (Plotly CDN + embedded images + hardcoded data).
They are generated in ddcm/docs/ and copied here for publication.
The workflow that produced them is recorded in REPORTING.md in the private
remote_main vault; it is not published here.
Unlinked snapshots
Committed 2026-08-04 but never added to any index. Kept for the record.
| Report | Topic |
|---|---|
| behavioral_validation_20260624.html through behavioral_validation_20260719.html | Nine behavioural-validation snapshots, 24 Jun – 19 Jul 2026, including 20260714_baseline |
| experiments_exact_sa_rl_20260624.html | Second Exact/SA/RL experiment run |
| recovery_K10_freemu_20260627.html | K=10 recovery, free μ |
| recovery_cell2_freemode_robust_20260702.html | Cell-2 free-mode robustness recovery |