=== RESEARCH ===
{'issue': 299, 'kind': 'docs', 'defect_status': 'confirmed', 'category': 'new-page', 'target_files': ['docs/src/notation.md', 'docs/src/bibliography.md', 'docs/src/index.md', 'docs/make.jl', 'docs/src/manual.md', 'docs/src/vecm.md', 'docs/src/lp.md', 'docs/src/structural_identification.md', 'docs/docrule.md'], 'summary': 'Create canonical docs/src/notation.md and docs/src/bibliography.md (extracted from index.md), register both under a Project/ nav section, and unify all 50 per-page reference lists to one Chicago author-date, en-dash, DOI-resolving canonical form.', 'current_state': 'Neither target page exists (ls docs/src/notation.md docs/src/bibliography.md -> "No such file or directory"). The extraction source is still wrong and unfixed: index.md:168 gives `u_t` as ``n \\times n`` (must be n×1 — every other page already agrees, e.g. id_heteroskedastic.md:104 and nongaussian.md:83 both say ``u_t`` is ``n \\times 1``). The notation table lives only in index.md:159-170. Bibliography is a per-page silo (50 pages carry `## References`) with concrete conflicts still present: (a) Blanchard-Quah 1989 appears at index.md:199 (Chicago style, NO DOI), manual.md:699-700 (APA style, DOI `10.3386/w2737` — the NBER working-paper DOI the issue explicitly forbids, currently live), lp.md:954 and structural_identification.md:354 (APA, no DOI); (b) Lütkepohl 2005 appears three ways — index.md:202 "Lutkepohl" no umlaut + ISBN 978-3-540-40172-8, manual.md:712 "Lütkepohl" + DOI 10.1007/978-3-540-27752-1, vecm.md:472 "Lutkepohl" no locator; (c) Sims 2002 at index.md:270 "20 (1--2)". Style is split site-wide: index.md uses Chicago author-date with `--` en-dash ranges ("655--673"), manual.md et al. use APA with `&`/parenthetical-year/hyphen ranges ("655-673"). z_t symbol clash is real: dsge_nonlinear.md:91 defines ``z_t`` as the n×1 endogenous vector, while dsge_continuous.md:42-52 uses ``z`` as labor-productivity/income state. make.jl:24-105 pages= has NO Project/ section. index.md @contents (line 307) omits notation.md/bibliography.md (and already omits x13, spectral, structural_identification, io_*, id_* pages — that broader trim is T182/#281\'s). docrule.md §References (311-333) still prescribes the per-page-silo model ("Full reference list at the bottom of every page (not just a master list)") and has no bijection/canonical/Chicago/en-dash rule and no new-page checklist — those docrule rules are added by T182/#281 and T196/#295, which this issue depends on.', 'proposed_change': '1. CREATE docs/src/notation.md: copy the notation table from index.md:159-170 verbatim EXCEPT change the `u_t` row from ``n \\times n`` to ``n \\times 1`` (source of truth: id_heteroskedastic.md:104, nongaussian.md:83, manual.md:267 all treat u_t as the n×1 residual vector). Add rows for the symbols the docrule dictionary (docrule.md:232-247) standardizes but index.md omits (x_t, theta, n, n_x, n_y, phi, sigma) so notation.md is the superset canonical dictionary. Resolve the z_t clash by giving ``z_t`` exactly one canonical meaning = "n×1 vector of endogenous variables (higher-order/nonlinear perturbation)" per dsge_nonlinear.md:91, and add an explicit note that dsge_continuous.md locally overrides ``z`` to mean the two-state labor-productivity/income process (dsge_continuous.md:42-52); cross-reference that local override rather than repurposing the dictionary symbol. Page is prose+tables only (no @example blocks needed).\n2. CREATE docs/src/bibliography.md: one canonical Chicago author-date entry per work, alphabetical by author, `--` en-dash page ranges, DOI links that resolve to the PUBLISHED article. Seed it from index.md:172-294 (already Chicago style) as the base, then fold in every work cited on child pages. Resolve the three named conflicts: (a) Blanchard, Olivier Jean, and Danny Quah. 1989. "The Dynamic Effects of Aggregate Demand and Supply Disturbances." *American Economic Review* 79 (4): 655--673. — use the AER article DOI 10.1257/... form resolved by DOI lookup (NOT `10.3386/w2737`); the implementer MUST resolve the DOI in a browser, not pattern-match. (b) Lütkepohl, Helmut. 2005. *New Introduction to Multiple Time Series Analysis*. Berlin: Springer. — one entry, correct umlaut, ONE locator (pick DOI 10.1007/978-3-540-27752-1 per manual.md:712 or the ISBN per index.md:202, not both). (c) Sims, Christopher A. 2002. — one volume(issue), keep "20 (1--2): 1--20" per index.md:270. Enter the Gregory-Hansen entry UNCHANGED from tests_breaks.md:609 including DOI 10.1016/0304-4076(69)41685-7 (verified legacy PII DOI — do not re-flag). Add the orphan/missing entries that T193/#292 and T194/#293 flag so they have a canonical home.\n3. UNIFY per-page reference lists to the canonical Chicago form: convert the APA-styled lists (manual.md:695-712+, vecm.md:472, lp.md:954, structural_identification.md:354, and every other page whose `## References` uses `&`/`(Year)`/hyphen ranges) to match bibliography.md\'s entry for each work. Per-page lists stay short (only works cited on that page) but must be byte-for-byte the canonical form. Critically, replace the forbidden NBER DOI at manual.md:700 with the AER DOI.\n4. index.md: replace the inline notation table (159-170) with a one-line pointer to notation.md (coordinate the exact trim with T182/#281 and T196/#295); add "notation.md" and "bibliography.md" to the @contents Pages array at index.md:307 in the SAME commit.\n5. make.jl: add a new top-level nav entry (created/owned by T196/#295) `"Project" => ["Notation" => "notation.md", "Bibliography" => "bibliography.md"]` to the pages= vector (currently ends at api_functions.md, line 104). No other make.jl cross-reference moves are required since these are brand-new pages.\n6. docrule.md: this issue assumes T182/#281 + T196/#295 have already added the §References bijection/canonical/style rules, the §Mathematics dictionary z_t fix, the Project nav, and the new-page checklist. If sequenced before them, the canonical/bijection/en-dash rules in docrule §References (311-333) must be added here too (replace the "Full reference list at the bottom of every page (not just a master list)" wording, which contradicts the canonical-bibliography model).', 'verify': "Primary: `julia --project=docs docs/make.jl` must build clean with the two new pages registered (warnonly is limited to :missing_docs and :cross_references per make.jl:110, so a broken @ref to notation.md/bibliography.md will NOT fail the build — manually confirm the Project section renders and every page's inline citation has a matching bibliography.md entry). notation.md and bibliography.md contain no @example/@setup blocks, so `julia --project=docs docs/verify_examples.jl docs/src/notation.md` and `.../bibliography.md` are trivial no-ops; run `julia --project=docs docs/verify_examples.jl docs/src/manual.md` (and vecm.md, lp.md, structural_identification.md) to confirm the reference-list rewrites did not disturb the example blocks on the pages you edited.", 'dependencies': [281, 295], 'collisions': ['docs/src/index.md — #281 (T182 fixes u_t at :168, trims @contents at :307, refreshes counts — same notation table and @contents this issue extracts/edits; hard dependency)', 'docs/make.jl — #295 (T196 creates the Project/ nav section this issue registers the two pages under)', "docs/src/manual.md — #292 (T193 LOW polish of multivariate pages incl. reference lists), #291 (T192 'ref-style split' sweep)", 'docs/src/vecm.md — #292 (T193 reference-list polish)', 'docs/src/lp.md — #292 (T193 reference-list polish)', 'docs/src/structural_identification.md — #293 (T194 identification-page polish)', 'docs/docrule.md — #281 (T182 §Mathematics z_t/u_t), #295 (T196 nav + new-page checklist), #304 (T205 citation-bijection lint reads this page as source of truth)'], 'effort': 'L', 'risk': 'med'}

=== VERIFY ===
{'issue': 299, 'verdict': 'CORRECTED', 'assessment': "Independently verified every cited file at the current release/v0.6.6 tip. The defect is real and largely as described: notation.md and bibliography.md do not exist; index.md:168 wrongly gives u_t as n×n while id_heteroskedastic.md:104 and nongaussian.md:82 (and manual.md:266-267 semantically) treat it as n×1; the notation table is at index.md:159-170; the forbidden NBER DOI 10.3386/w2737 is live at manual.md:700; the z_t clash is genuine (dsge_nonlinear.md:91 n×1 endogenous vector vs dsge_continuous.md:42-52 z = labor-productivity state — spec's line 91 cite is exact); make.jl has no Project section (pages= ends api_functions.md:104, warnonly line 110); index.md @contents:307 omits both pages; docrule.md:311-333 still prescribes the per-page-silo APA model and docrule.md:232-247 dictionary lacks z_t; Sims 2002 at index.md:270 = '20 (1--2): 1--20'; 50 pages carry ## References. The fix is directionally correct and obeys docrule's intent, but has two material problems. (A) Completeness gap: the plan extracts index.md:172-294 into bibliography.md but never says to remove/pointer-ize that MASTER topic-organized list, which would leave two competing master bibliographies. (B) The Blanchard-Quah DOI guidance hardcodes an assumed '10.1257/... form'; no such DOI is in-repo, whereas lp.md:955 already carries 10.2307/1827924 and structural_identification.md:355 the JSTOR-URL form of the same stable id, both resolving to the published AER article — assuming a 10.1257 prefix risks fabricating a DOI, contra docrule. Several current_state labels are also stale (lp/SI are not 'no DOI'; vecm.md:472-473 has ISBN, not 'no locator'; Lütkepohl appears in ~11 pages, not three) and the implementer must rely on current file state.", 'corrections': "1. index.md master list: after seeding bibliography.md from index.md:172-294, REPLACE index.md:172-294 with a one-line pointer to bibliography.md (mirroring the notation-table pointer in step 4); otherwise two master bibliographies coexist. Coordinate this trim with #281's @contents/count refresh in the same commit. 2. Blanchard-Quah DOI: do NOT assume a '10.1257/...' prefix. The canonical published-article DOI must be verified by lookup; 10.2307/1827924 (already used at lp.md:955, and the JSTOR-URL form at structural_identification.md:355) resolves to the published AER article and is an acceptable canonical choice — the only hard requirement is replacing the NBER 10.3386/w2737 at manual.md:700. Obey docrule 'do not fabricate or guess DOIs.' 3. Ignore stale current_state descriptors: lp.md:955 has DOI 10.2307/1827924 (not 'no DOI'); structural_identification.md:355 has a JSTOR URL (not 'no DOI'); vecm.md:472-473 has ISBN 978-3-540-40172-8 (not 'no locator'); Lütkepohl 2005 appears across ~11 pages (ia_irf, tests_diagnostics, innovation_accounting, id_testing, ia_fevd, tests, vecm, ia_hd, nongaussian, index, manual) with mixed umlaut/style/locator — unify all, not just three. 4. Sequencing: Chicago-author-date conversion contradicts the current docrule §References APA/per-page-silo template; do NOT convert pages until docrule §References is rewritten (via #281/#295 or step 6 here) or the pages will violate the standing docrule. 5. Adding z_t to notation.md overlaps semantically with y_t (docrule dict: y_t = endogenous variables); keep z_t explicitly scoped to the higher-order/perturbation stacked vector to avoid a new y_t/z_t ambiguity.", 'collision_issues': [281, 295, 291, 292, 293, 304], 'discriminating_power': 'n/a — docs-only issue. notation.md/bibliography.md are prose+tables (no @example/@setup), so their verify_examples runs are no-ops; the only test surface is re-running verify_examples on manual.md/vecm.md/lp.md/structural_identification.md after the reference-list rewrites, and those edits are at page bottoms and do not touch example blocks (low risk).', 'final_effort': 'L'}