data-analysis
lingzhi227/agent-research-skills
Generate rigorous statistical analysis code with multi-round review and proper uncertainty reporting.
What is data-analysis?
Produces statistical analysis code with 4-round peer review for experimental data. Selects appropriate tests based on data type, reports p-values and effect sizes, and generates analysis reports with confidence intervals. Use when analyzing experimental results for papers or research.
- Detects data types and recommends appropriate statistical tests (t-tests, ANOVA, Mann-Whitney U, Chi-square, correlation, regression)
- Generates analysis code with structured sections: imports, data loading, preprocessing, descriptive statistics, and hypothesis testing
- Performs 4-round code review checking for mathematical errors, data handling issues, sensible values, and cross-table consistency
- Formats p-values with significance stars, LaTeX notation, or plain text for publication
- Calculates effect sizes, confidence intervals, and standard deviations for all nominal values
- Handles missing values, unit conversions, and confounding variables in analysis
How to install data-analysis
npx skills add https://github.com/lingzhi227/agent-research-skills --skill data-analysis- Python with numpy, scipy, statsmodels, sklearn, and pandas installed
- Data in CSV, JSON, pickle, or experiment log format
- Clear research goal or hypothesis to test
How to use data-analysis
- 1.Provide your data source (CSV, JSON, pickle, or experiment logs) and your research goal or hypothesis
- 2.The skill generates analysis code structured with imports, data loading, preprocessing, and statistical tests appropriate for your data type
- 3.Review the generated code through the 4-round review process: check for mathematical errors, data handling issues, value sensibility, and cross-table consistency
- 4.Run the analysis code to produce results with p-values, effect sizes, and confidence intervals
- 5.Format p-values for publication using the format-pvalue script (stars, LaTeX, or plain text)
Use cases
- Analyzing experimental results from A/B tests or method comparisons to determine statistical significance
- Processing survey or observational data with multiple groups to test hypotheses using appropriate tests
- Generating publication-ready statistical summaries with p-values and effect sizes for research papers
- Comparing treatment groups in clinical or behavioral studies with proper uncertainty quantification
- Validating machine learning model performance across multiple runs with statistical rigor
- Researchers and academics analyzing experimental data
- Data scientists validating model performance with statistical tests
- Graduate students preparing statistical analyses for theses or papers
- Anyone conducting hypothesis testing on experimental or observational data
data-analysis FAQ
It supports t-tests, paired t-tests, Mann-Whitney U, ANOVA, Kruskal-Wallis, Chi-square, Fisher's exact, Pearson/Spearman correlation, OLS regression, logistic regression, and mixed-effects models. Test selection is automatic based on data type and sample characteristics.
Yes. The skill includes data handling review in Round 2 of the 4-round review process, checking for missing values, units, preprocessing, and appropriate test choice for your data.
Yes. Every nominal value must have uncertainty (confidence interval, standard deviation, or p-value), and the skill enforces this. Results must match actual data—no hallucination.
pandas, numpy, scipy, statsmodels, and sklearn. All are standard Python data science libraries.
Round 1 checks for mathematical/statistical errors; Round 2 reviews data handling and test choice; Round 3 validates individual table values and uncertainty; Round 4 ensures cross-table consistency and completeness.
Full instructions (SKILL.md)
Source of truth, from lingzhi227/agent-research-skills.
name: data-analysis description: Generate statistical analysis code with 4-round review. Select appropriate statistical tests, interpret results, and produce analysis reports with p-values, effect sizes, and confidence intervals. Use when analyzing experimental data for a paper. argument-hint: [data-source]
Data Analysis
Generate rigorous statistical analysis code with multi-round review.
Input
$0— Data source (CSV, JSON, pickle, or experiment logs)$1— Research goal or hypothesis to test
References
- 4-round code review prompts:
~/.claude/skills/data-analysis/references/review-prompts.md
Scripts
Statistical summary and comparison
python ~/.claude/skills/data-analysis/scripts/stat_summary.py --input results.csv --compare method --metric accuracy --output summary.json
python ~/.claude/skills/data-analysis/scripts/stat_summary.py --input results.csv --describe
Detects data types, recommends tests, runs comparisons, outputs effect sizes and significance stars. Requires numpy, scipy.
Format p-values
python ~/.claude/skills/data-analysis/scripts/format_pvalue.py --values "0.001 0.05 0.23" --format stars
python ~/.claude/skills/data-analysis/scripts/format_pvalue.py --csv results.csv --column pvalue --format latex
Formats p-values with stars, LaTeX notation, or plain text. Stdlib-only.
Workflow
Step 1: Generate Analysis Code
Structure the code with these sections:
# IMPORT— pandas, numpy, scipy, statsmodels, sklearn# LOAD DATA— Load from original data files# DATASET PREPARATIONS— Missing values, units, exclusion criteria# DESCRIPTIVE STATISTICS— Summary tables if needed# PREPROCESSING— Dummy variables, normalization# ANALYSIS— Statistical tests per hypothesis# SAVE ADDITIONAL RESULTS— Extra results to pickle
Step 2: 4-Round Code Review
- Round 1 — Code Flaws: Mathematical/statistical errors, wrong calculations, trivial tests
- Round 2 — Data Handling: Missing values, units, preprocessing, test choice
- Round 3 — Per-Table: Sensible values, measures of uncertainty, missing data
- Round 4 — Cross-Table: Completeness, consistency, missing variables
Step 3: Produce Results
- Every nominal value must have uncertainty (CI, STD, or p-value)
- Statistical tests must be appropriate for the data type
- Results must match actual data — never hallucinate
Allowed Packages
pandas, numpy, scipy, statsmodels, sklearn, pickle
Statistical Test Selection
| Data Type | Test |
|---|---|
| Two groups, normal | Independent t-test |
| Two groups, non-normal | Mann-Whitney U |
| Paired samples | Paired t-test / Wilcoxon |
| Multiple groups | ANOVA / Kruskal-Wallis |
| Categorical | Chi-square / Fisher's exact |
| Correlation | Pearson / Spearman |
| Regression | OLS / Logistic / Mixed effects |
Rules
- Always report p-values for statistical tests
- Account for relevant confounding variables
- Use inherent package functionality (e.g.,
formula = "y ~ a * b"for interactions) - Do not manually implement available statistical functions
- Access dataframes using string-based column names, not integer indices
Related Skills
- Upstream: experiment-code, experiment-design
- Downstream: table-generation, figure-generation, backward-traceability
- See also: math-reasoning
Related skills
More from lingzhi227/agent-research-skills and the wider catalog.
deep-research
Conduct systematic 6-phase academic literature reviews with structured output and synthesis.
excalidraw-skill
Programmatic canvas toolkit for creating and refining Excalidraw diagrams with real-time sync and element-level control.
experiment-code
Generate and iteratively improve ML experiment code for research papers with training pipelines, debugging, and result optimization.
experiment-design
Design structured, progressive experiment plans for research papers with staged baselines, hyperparameter sweeps, and ablation studies.
figure-generation
Generate publication-quality scientific figures with matplotlib/seaborn using query expansion, code generation, and VLM visual feedback.
github-research
Discover and analyze GitHub repositories for research topics using deep-research output and multi-source search.