clarify respondent packet scoring rows

Co-Authored-By: PI[gpt-5.6-terra] <288921227+claudypoo@users.noreply.github.com>

docs/RESEARCH_JOURNAL.md

@@ -1286,6 +1286,12 @@ question instances because two question names are paraphrases rather than visibl
 prompt IDs. Twenty packets have too-long reasons but retain their selected answers; reasons were
 not scored.
 
+Coordinate comparison note: both old and new analyses score the same 12 recovered WVS axis rows,
+8 ordinary items plus the four `PANEL_QUALITIES` child rows, Obedience, Independence,
+Determination/perseverance, and Imagination. The packet asks all 11 human child qualities. The
+other seven selections, including Religious faith, remain in raw respondent packets but do not enter
+these coordinates because no axis coefficients are used for them here.
+
 Observation, from the fixed whole-row bootstrap in
 `slop/research/wvs/20260919_respondent_packet/analysis.json`: packet constant-family RMSE is
 0.0792 and linear RMSE is 0.0696. Constant LOO error is 0.0942, lower than linear LOO error 0.1318.
@@ -1296,4 +1302,4 @@ a causal model-release effect. The published WVS map data, PNG, and SVG had a by
 
 Full audit: `slop/audits/job_1788_wvs_respondent_packet_v2.md`.
 
--- PI[unknown]
+-- PI[gpt-5.6-terra]

slop/audits/job_1788_wvs_respondent_packet_v2.md

@@ -65,7 +65,7 @@ A fresh read-only reviewer separately audited every raw structure and identified
 
 ## Reanalysis
 
-Coordinates are `[Survival <-> Self-expression, Traditional <-> Secular-Rational]`. Packet CIs and SEs use B=1,000 whole-respondent-row bootstrap draws. Dense-v1 uses its existing per-item rating samples, so packet-minus-dense shifts use independent bootstrap draws.
+Coordinates are `[Survival <-> Self-expression, Traditional <-> Secular-Rational]`. Both comparisons score the same 12 recovered WVS axis rows: 8 ordinary items plus the four child-quality coefficient rows in `PANEL_QUALITIES`, Obedience, Independence, Determination/perseverance, and Imagination. The packet asks all 11 human child qualities. The other seven selections, including Religious faith, remain in every raw respondent packet but do not enter these recovered coordinates because no axis coefficients are used for them here. Packet CIs and SEs use B=1,000 whole-respondent-row bootstrap draws. Dense-v1 uses its existing per-item rating samples, so packet-minus-dense shifts use independent bootstrap draws.
 
 | release | packet coordinate, SE | packet CI95 | dense-v1 coordinate, SE | packet minus dense, CI95 |
 |---|---|---|---|---|
@@ -214,4 +214,4 @@ Two full deterministic analysis executions produced the same SHA-256s:
 
 Map diff exit code was 0. Map SHA-256 values are `c03b5104a9a31482fa547a8a2996e33d3d3d59453d5189f94e95d9d451bf63d1` for JSON, `a7164901ff69e1e51696ade8fb9dd9c62d81af5cfa0c94a49ddf5598c68b4103` for PNG, and `d35cb80a70d34472e967ac1453aa54d43c24418b8b74fe71e726d2a927771aa2` for SVG.
 
--- PI[unknown]
+-- PI[gpt-5.6-terra]