docs(community): add SUPPORT.md and GitHub issue templates

JOSS reviewer checklist requires clear pathways for: (1) contributing,
(2) reporting issues, (3) seeking support. CONTRIBUTING.md and
CODE_OF_CONDUCT.md already covered (1) and conduct; this commit fills
in the issue-reporting and support pathways.

Added:
- SUPPORT.md: where to ask usage questions (Discussions), where to file
  bugs (Issues + bug_report template), how to report security issues
  privately (email), and response-time expectations.
- .github/ISSUE_TEMPLATE/bug_report.md: structured form including
  version probe, minimal-reproducible-example block keyed to bundled
  sp.datasets, and a dedicated section for numerical-correctness reports
  against Stata / R / source-paper reference implementations.
- .github/ISSUE_TEMPLATE/feature_request.md: structured form including
  proposed API sketch and a required References section (DOI / arXiv ID
  for new estimators, since this project's correctness policy forbids
  unverified citations).
- .github/ISSUE_TEMPLATE/config.yml: disable blank issues; route
  how-do-I questions to Discussions and security reports to private
  email instead of the public tracker.

Pre-review smoke test confirmed the README Quick Start exercises
(DiD callaway_santanna on mpdta, IV ivreg on card_1995, RD rdrobust on
lee_2008_senate, Synth on california_prop99) all run end-to-end in a
fresh venv via `pip install statspai` and return numbers matching the
README's stated expectations.

Author: brycewang-stanford

.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,59 @@
+---
+name: Bug report
+about: Report a crash, incorrect number, documentation error, or behavior that contradicts the docstring/paper
+title: "[bug] <short description>"
+labels: bug
+assignees: ''
+---
+
+## Summary
+
+<!-- One sentence: what is wrong. -->
+
+## Environment
+
+- StatsPAI version: <!-- python -c "import statspai as sp; print(sp.__version__)" -->
+- Python version: <!-- python --version -->
+- OS: <!-- macOS / Linux / Windows + version -->
+- Install source: <!-- pip install statspai / pip install -e . / conda / source build -->
+- Extras installed (if any): <!-- e.g. [plotting], [fixest], [neural], [performance], [bayes] -->
+
+## Minimal reproducible example
+
+<!--
+If at all possible, please use one of the bundled teaching datasets under
+`sp.datasets` (mpdta, card_1995, lee_2008_senate, california_prop99, ...)
+so we can run the example directly.
+-->
+
+```python
+import statspai as sp
+
+# your minimal example here
+```
+
+## What you expected
+
+<!-- The number / behavior you expected. Reference the docstring, paper,
+or reference implementation (Stata / R / other) if applicable. -->
+
+## What actually happened
+
+<!-- Paste the error traceback or the unexpected output. -->
+
+```
+<traceback or output here>
+```
+
+## Numerical-correctness notes (if applicable)
+
+<!-- For estimator-level numerical issues, fill in:
+- Reference implementation tried for comparison (Stata package / R package / source paper)
+- Version of that reference (Stata 18, fixest 0.11, did 2.1.2, etc.)
+- The number it returns vs. what StatsPAI returns
+- A reproducible script if possible
+-->
+
+## Anything else
+
+<!-- Logs, screenshots, related issues, suspected cause. -->

.github/ISSUE_TEMPLATE/config.yml

@@ -0,0 +1,8 @@
+blank_issues_enabled: false
+contact_links:
+  - name: Usage question / how-do-I
+    url: https://github.com/brycewang-stanford/StatsPAI/discussions
+    about: For "how do I do X", methodological discussion, or comparing results across Stata/R/StatsPAI — please use GitHub Discussions.
+  - name: Security vulnerability
+    url: mailto:brycew6m@stanford.edu
+    about: Please report security issues privately via email, not as a public issue.

.github/ISSUE_TEMPLATE/feature_request.md

@@ -0,0 +1,44 @@
+---
+name: Feature request / new estimator
+about: Suggest a new feature, estimator, or improvement
+title: "[feature] <short description>"
+labels: enhancement
+assignees: ''
+---
+
+## What you would like
+
+<!-- One paragraph: the feature, estimator, or improvement you're proposing. -->
+
+## Motivation
+
+<!-- Why is this useful? What problem does it solve? Who benefits? -->
+
+## Proposed API (if applicable)
+
+<!-- A sketch of what calling this would look like. Match StatsPAI's existing
+`sp.<verb>(data=..., ...)` style where possible. -->
+
+```python
+import statspai as sp
+
+# proposed usage
+```
+
+## References
+
+<!-- For new estimators, this section is required:
+- Source paper(s) with DOI or arXiv ID
+- Existing reference implementations (Stata package, R package, etc.)
+- Any benchmarks or canonical datasets the estimator is usually tested on
+-->
+
+## Alternatives considered
+
+<!-- What other approaches did you consider? Are there existing functions in
+StatsPAI (or Stata / R) that partially address this? -->
+
+## Anything else
+
+<!-- Notes on numerical pitfalls, edge cases, scope, or anything else
+that would help us scope the work. -->

SUPPORT.md

@@ -0,0 +1,45 @@
+# Getting Support for StatsPAI
+
+Thanks for using StatsPAI. Pick the channel that fits what you need.
+
+## Asking a question or discussing usage
+
+For "**how do I**", "**which estimator should I use**", "**why does my result differ from Stata / R**", or general usage discussion:
+
+- **GitHub Discussions** — <https://github.com/brycewang-stanford/StatsPAI/discussions>
+  - Best place for open-ended questions, methodological discussion, and Q&A that may help future users.
+  - Please search existing threads first — many common questions are already answered.
+
+When asking, please include:
+
+1. StatsPAI version: `python -c "import statspai as sp; print(sp.__version__)"`
+2. Python version + OS
+3. Minimal reproducible example (data + code, ideally using one of the bundled `sp.datasets`)
+4. What you expected vs. what you got
+
+## Reporting a bug
+
+If something is **wrong** — a crash, an incorrect number, a documentation error, or behavior that contradicts the docstring / paper:
+
+- **GitHub Issues** — <https://github.com/brycewang-stanford/StatsPAI/issues>
+- Use the **Bug report** template.
+- Numerical-correctness issues (estimator returns a different value than Stata / R / the source paper) are top priority. Please cite the reference implementation and version when relevant.
+
+## Requesting a feature or new estimator
+
+- **GitHub Issues** — use the **Feature request / new estimator** template.
+- For new estimators, please link to the source paper (with DOI) and, if possible, a reference implementation in Stata, R, or another package.
+
+## Security issues
+
+If you believe you have found a security vulnerability (e.g., a code path that allows arbitrary code execution from data input), please **do not** open a public issue. Email the maintainer directly:
+
+- **Email:** brycew6m@stanford.edu
+
+## Commercial support and collaboration
+
+StatsPAI is developed by the team behind [CoPaper.AI](https://copaper.ai) (Stanford REAP). For applied-econometrics consulting, custom-estimator collaborations, or institutional licensing questions, email brycew6m@stanford.edu.
+
+## Response time expectations
+
+StatsPAI is maintained primarily by a small team. We aim to triage issues within 1–2 weeks. For urgent numerical-correctness bugs or breakage in core estimators (`did`, `iv`, `rd`, `synth`, `dml`, `panel`), we will respond faster.