Robustness Table
Testing & Qualityrobustness-tableRobustness Table is an Agent Skill for testing & quality published in the brycewang-stanford/Auto-Empirical-Research-Skills repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Generates robustness check code and formats results as a combined table. Use for sensitivity analysis.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/HEAD/skills/29-quarcs-lab-project20XXy/dot-claude/skills/robustness-table/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/robustness-table, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Generate Robustness Checks and Table
Given a baseline regression, generate code for standard robustness checks and format the results as a publication-ready table.
Arguments
$ARGUMENTS— notebook reference and baseline specification description (e.g., "notebook-02 baseline OLS with GDP on life expectancy")
Steps
-
Read the specified notebook and locate the baseline regression:
- Look for estimation commands (Python:
statsmodels,linearmodels; R:lm,fixest,felm; Stata:reg,reghdfe,ivregress) - Identify the dependent variable, independent variables, fixed effects, and clustering
- Look for estimation commands (Python:
-
Ask the user which robustness checks to include:
- Alternative control variable sets (drop/add controls)
- Alternative fixed effects specifications
- Different standard error clustering levels
- Subsample analysis (e.g., by region, time period, income group)
- Winsorized or trimmed dependent variable
- Alternative dependent variable (e.g., log vs level)
- Placebo tests (randomized treatment, pre-period outcome)
- Alternative estimation methods (e.g., OLS vs Poisson, logit vs probit)
-
Generate code cells in the notebook for each robustness specification:
- Each cell should be self-contained (loads data, runs regression, stores results)
- Use consistent variable naming for results collection
-
Create a summary cell that collects all results into a single table:
- Cell directive:
#| label: tbl-robustness(or*|for Stata) - Cell directive:
#| tbl-cap: "Robustness checks"(or*|for Stata) - Format: baseline in column (1), each robustness check in subsequent columns
- Follow academic conventions: coefficient (SE), significance stars, N, R², FE indicators
- Stata caveat: Do NOT use
tbl-prefix for Stata text output — use a plain label (e.g.,stata-robustness)
- Cell directive:
-
Optionally export the table to
tables/as a standalone file (LaTeX or CSV) -
Sync the Jupytext pair:
uv run jupytext --sync notebooks/<name>.md -
Show the embed shortcode for
index.qmd:{{< embed notebooks/<name>.ipynb#tbl-robustness >}}
Error handling
- If the baseline regression is not found, ask the user to point to the specific cell.
- If the notebook uses a language not recognized, ask for guidance on the estimation syntax.