Data Analysis Workflow

Run an end-to-end data analysis in R or Python: load, explore, analyze, and produce publication-ready output.

Input: $ARGUMENTS — a dataset path (e.g., data/county_panel.csv) or a description of the analysis goal (e.g., "regress wages on education with state fixed effects using CPS data").

Before making analysis changes, read ${CLAUDE_PLUGIN_ROOT}/rules/orchestrator-protocol.md, ${CLAUDE_PLUGIN_ROOT}/rules/analysis-verification.md, and ${CLAUDE_PLUGIN_ROOT}/rules/quality-gates.md; for replication or extension requests, also read ${CLAUDE_PLUGIN_ROOT}/rules/replication-protocol.md.

Phase 0: Choose Language

R is the default. If ambiguous, use AskUserQuestion to pick R (Recommended — tidyverse/fixest/ggplot2, full plugin support including r-reviewer) or Python (pandas/statsmodels). Pick R when the project has existing .R code; Python when it has existing .py or .ipynb.

Phase 0.5: Ingest Non-Tabular Inputs

If $ARGUMENTS points to a PDF, HTML page, government portal URL, or other non-tabular source, Read ${CLAUDE_PLUGIN_ROOT}/skills/data-analysis/references/ingest-recipes.md and follow it before Phase 1. Skip this phase for flat files (.csv, .parquet, .dta, .rds, .feather, .json).

R Track

Constraints

• Follow ${CLAUDE_PLUGIN_ROOT}/rules/r-code-conventions.md for all standards
• Save scripts to scripts/R/ with descriptive names
• Save all outputs (figures, tables, RDS) to output/
• Use saveRDS() for every computed object
• Run r-reviewer on the generated script before presenting results

Phase 1: Setup and Data Loading

1. Create R script with proper header (title, author, purpose, inputs, outputs) — Read ${CLAUDE_PLUGIN_ROOT}/skills/data-analysis/assets/analysis-script.R and adapt
2. Load required packages at top (library(), never require())
3. Set seed once at top: set.seed(42)
4. Create output directories: dir.create("output/analysis", recursive = TRUE, showWarnings = FALSE)
5. Load and inspect the dataset

Phase 2: Exploratory Data Analysis

• summary(), missingness rates, variable types
• Histograms for key continuous variables
• Scatter plots, correlation matrices
• Panel trends, pre-treatment comparisons if applicable
• Save all diagnostic figures to output/diagnostics/

Phase 3: Main Analysis

• Panel data: use fixest; cross-section: use lm/glm
• Cluster SEs at the appropriate level (document why)
• Multiple specifications: start simple, progressively add controls
• Report standardized effects alongside raw coefficients

Phase 3.5: Identification Diagnostics

Mandatory for any analysis with a causal interpretation. Detect the design from the analysis goal and the model call (feols(... | id + t) with treatment-timing → DiD; rdrobust → RDD; feols(y ~ ... | 0 | endog ~ instr) → IV; tidysynth → synthetic control; random assignment → RCT; MatchIt/cobalt → matching). Read ${CLAUDE_PLUGIN_ROOT}/skills/data-analysis/references/identification-r.md and run the block matching the detected design. Ask once with AskUserQuestion only if the design is genuinely ambiguous. Skip this phase for purely descriptive analyses.

Phase 4: Publication-Ready Output

Tables: modelsummary (preferred) or stargazer — export .tex and .html Figures: ggplot2 with project theme; explicit ggsave(width = X, height = Y); save as .pdf and .png; add bg = "transparent" only if output is for Beamer slides

Phase 5: Save and Review

1. saveRDS() for all key objects
2. Dispatch the r-reviewer agent via Task:

   Task prompt: "Review the R script at scripts/R/[script_name].R against
   ${CLAUDE_PLUGIN_ROOT}/rules/r-code-conventions.md. Save your report to
   quality_reports/[script_name]_r_review.md. Follow the r-reviewer agent instructions."

3. Address Critical and High issues before presenting results

Python Track

Constraints

• Save scripts to scripts/python/ with descriptive names
• Save all outputs (figures, tables, pickles) to output/
• Use joblib.dump() for model objects; .to_parquet() for DataFrames
• Use pathlib.Path for all file paths — never hardcode absolute paths
• Set random seeds at the top of the script

Phase 1: Setup and Data Loading

1. Create Python script with header — Read ${CLAUDE_PLUGIN_ROOT}/skills/data-analysis/assets/analysis-script.py and adapt
2. Import all packages at the top of the file
3. Set seeds: np.random.seed(42) and random.seed(42)
4. Create output directories: Path("output/analysis").mkdir(parents=True, exist_ok=True)
5. Load and inspect the dataset with pandas

Phase 2: Exploratory Data Analysis

• df.describe(), df.isnull().sum(), df.dtypes
• Histograms and distributions with matplotlib/seaborn
• Scatter plots and correlation matrices
• Save diagnostic figures to output/diagnostics/
• Save summary stats: df.describe().to_csv("output/diagnostics/summary_stats.csv")

Phase 3: Main Analysis

• Cross-section OLS: smf.ols("y ~ x", data=df).fit(cov_type="HC3")
• Panel data: PanelOLS from linearmodels with cluster-robust SEs
• Multiple specifications: build incrementally
• Document SE choice with a comment

Phase 3.5: Identification Diagnostics

Mandatory for any analysis with a causal interpretation. Detect the design from the analysis goal and the model call (PanelOLS(..., entity_effects=True, time_effects=True) with treatment timing → DiD; rdrobust port → RDD; IV2SLS → IV; pysyncon → synthetic control; experiment → RCT; causalinference → matching). Read ${CLAUDE_PLUGIN_ROOT}/skills/data-analysis/references/identification-py.md and run the block matching the detected design. Skip this phase for purely descriptive analyses.

Phase 4: Publication-Ready Output

Tables: Format with pandas and export via .to_latex() or stargazer (Python port) Figures: matplotlib/seaborn; explicit fig.savefig(path, dpi=300, bbox_inches="tight"); save as .pdf and .png

Phase 5: Save and Review

1. joblib.dump(model, "output/model.pkl") for fitted models
2. df_results.to_parquet("output/results.parquet") for DataFrames
3. Review manually against the checklist below before presenting

Python Quality Checklist

[ ] All imports at top
[ ] Random seeds set (numpy + stdlib)
[ ] All paths use pathlib.Path — no hardcoded strings
[ ] Output directories created with mkdir(exist_ok=True)
[ ] Figures saved with explicit dpi=300, bbox_inches="tight"
[ ] Model objects saved with joblib.dump()
[ ] DataFrames saved as parquet
[ ] Comments explain WHY, not WHAT

Shared Principles

• Reproduce, don't guess. If the user specifies a regression, run exactly that.
• Show your work. Compute summary statistics before jumping to regression.
• Check for issues. Look for multicollinearity, outliers, perfect prediction, missing data.
• Use relative paths. All paths relative to repository root.
• No hardcoded values. Use variables for sample restrictions, date ranges, thresholds.