feat(reporting): add full_report() six-stat bundler (n / CI / method / p / effect size / statistic)

[ywatanabe1989]

Summary

Driven by the operator's six-stat reporting doctrine (2026-07-05, tracked
as scitex-todo card scitex-stats-six-stat-report-doctrine): every
reported statistic must carry all SIX of (1) n, (2) 95% CI, (3)
method/test name, (4) p-value, (5) effect size, (6) test statistic.
Partial reporting is treated as incomplete.

Today's test_*() / run_test() result dicts already carry 5 of 6
(method, statistic, pvalue, effect_size, n) — there was no confidence-
interval field anywhere. This PR adds a full_report() helper that
bundles all six into one dict + a human-readable string, and makes
"missing a field" a raised error, not a silent gap.

• scitex_stats.reporting.full_report(result, *, data=None, data2=None, ci=None, confidence=0.95, n_bootstrap=10_000, random_state=None, strict=True)
  (src/scitex_stats/reporting/_full_report.py)
  • Accepts a run_test()/test_*()-style result dict, plus either an
    already-computed ci=(lower, upper) tuple or the raw data/data2
    arrays so it can derive one itself.
  • CI derivation: analytic closed-form via
    scipy.stats.ttest_ind/ttest_rel/ttest_1samp(...).confidence_interval()
    for parametric t-tests (reusing scipy's own machinery rather than
    re-deriving the formula); percentile bootstrap via
    scipy.stats.bootstrap for anything without a closed form. Prefers
    scitex_stats.resampling.bootstrap_ci when importable (that module
    is currently on open, unmerged PR feat(resampling): add auc_ci, delta_auc_ci, bootstrap_ci #66 — this hands off to it
    once it lands, rather than duplicating it now).
  • Raises IncompleteReportError (a ValueError subclass) when any of
    the six fields can't be determined — pass strict=False to log a
    warning and return a partial report with missing_fields instead.
  • Returns: method, statistic, stat_symbol, pvalue,
    effect_size, effect_size_metric, n, ci, ci_level,
    formatted, missing_fields.
• Wired into the top-level lazy loader (src/scitex_stats/__init__.py),
  following the exact _LAZY_ATTRS/__all__ pattern used for the other
  submodules (auto, descriptive, etc.) — import scitex_stats as ss; ss.full_report(...) works without eagerly importing scipy/matplotlib.
• Added fmt_sym_md() to _utils/_formatters.py: a markdown-italic
  counterpart to the existing matplotlib-mathtext fmt_sym(), for
  plain-text six-stat strings (e.g. *t*, *n*_x, *N*_subjects).
  Preserves the N (subject-level) / n (window-level) convention
  verbatim — it only italicizes the base letter, it doesn't decide N vs
  n (a generic array-based test function has no way to know which level
  its input represents; that judgement stays with the caller).
• Updated the module docstring "Functionalities" bullets in
  src/scitex_stats/__init__.py.

Example

import numpy as np
from scitex_stats import run_test, full_report

x = np.random.default_rng(0).normal(0, 1, 50)
y = np.random.default_rng(1).normal(0.8, 1, 50)
result = run_test("ttest_ind", data=x, data2=y)
report = full_report(result, data=x, data2=y)
print(report["formatted"])
# Welch's t-test (independent): *t* = 2.34, *p* = .021, *d* = 0.47,
# 95% CI [0.12, 0.89], *n*_x = 50, *n*_y = 50

Notes on scope

• No existing test_*() files were rewritten — this is additive. I
  spot-checked several _plot_* call sites (fmt_sym('n') usage across
  shapiro/pearson/kendall/wilcoxon/anova/ttest_rel) and found the
  existing matplotlib italics already correct; no bugs to spot-fix
  there.
• scitex_stats.resampling / effect_size_from_ci (referenced in the
  original task brief as "already landed") are actually still on open
  PR feat(resampling): add auc_ci, delta_auc_ci, bootstrap_ci #66, not merged into develop — verified via gh pr view 66
  (mergedAt: null). This PR's bootstrap fallback uses
  scipy.stats.bootstrap (already a hard dependency) directly instead,
  with a guarded try/except ImportError hand-off to
  scitex_stats.resampling.bootstrap_ci for if/when feat(resampling): add auc_ci, delta_auc_ci, bootstrap_ci #66 merges.

Test plan

• pytest tests/scitex_stats/reporting/ — 19 new tests, all pass
• pytest tests/scitex_stats/_utils/test__fmt_sym_md.py — 4 new tests, all pass
• pytest tests/scitex_stats/_utils/ tests/scitex_stats/test__dispatch.py — 249 passed, no regressions
• pytest tests/develop/test_audit.py (PA-307 audit_all_for_package) — passes; fixed 4 initial STX-TQ002 (AAA-marker) violations in the new test file before this
• scitex-dev linter check-files on all changed files — only pre-existing-pattern warnings (scipy import, matplotlib import — same as e.g. tests/parametric/_test_ttest_ind.py), no new errors

[github-actions]

Please sign the SciTeX CLA before your contribution can be merged.
Comment I have read and agree to the SciTeX CLA. to sign.

---

I have read the CLA Document and I hereby sign the CLA

---

_{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}