Data Analysis Workflow

Run an end-to-end data analysis in R: 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").

Constraints

• Follow R code conventions in .claude/rules/r-code-conventions.md
• Save all scripts to code/ with descriptive names
• Save all outputs (figures, tables, cached data) to output/
• Cache data with fst::write_fst() for tabular, qs2::qs_save() for model objects
• Use ggthemes::theme_stata() for all figures
• Run r-reviewer on the generated script before presenting results

Workflow Phases

Phase 0: Pre-Flight Report

Before writing any analysis code, produce a Pre-Flight Report showing you read the inputs. This prevents the common failure mode where the agent hallucinates variable names or skips project conventions.

Output block (in your response to the user, before Phase 1):

markdown
1## Pre-Flight Report
2
3**Dataset:** [path]
4- Variables found: [list from head()/names()]
5- Rows: [count]
6- Key types: [e.g., "outcome=numeric, treatment=binary, state=factor"]
7- Missing-data summary: [% missing per key var]
8
9**Project conventions read:**
10- `.claude/rules/r-code-conventions.md` — [one-line summary of most relevant rule]
11- `.claude/rules/content-invariants.md` — [INV-9, INV-10, INV-11, INV-12 applicable]
12
13**Task interpretation:** [one sentence restating what the user asked for]
14
15**Plan:** [3-5 bullet outline of the R script structure]

If any input cannot be read (missing file, unreadable format), stop and ask the user before proceeding.

Phase 1: Setup and Data Loading

1. Read .claude/rules/r-code-conventions.md for project standards
2. Create R script with proper header (title, author, purpose, inputs, outputs)
3. Load required packages at top via library() (renv + pak for install)
4. Call set.seed(888) before each step involving randomness (per r-code-conventions.md)
5. Load and inspect the dataset

Phase 2: Exploratory Data Analysis

Generate diagnostic outputs:

• Summary statistics: summary(), missingness rates, variable types
• Distributions: Histograms for key continuous variables
• Relationships: Scatter plots, correlation matrices
• Time patterns: If panel data, plot trends over time
• Group comparisons: If treatment/control, compare pre-treatment means

Save all diagnostic figures to output/diagnostics/.

Phase 3: Main Analysis

Based on the research question:

• Regression analysis: Use fixest for panel data, lm/glm for cross-section
• Standard errors: Cluster at the appropriate level (document why)
• Multiple specifications: Start simple, progressively add controls
• Effect sizes: Report standardized effects alongside raw coefficients

Phase 4: Publication-Ready Output

Tables:

• Use modelsummary for regression tables (preferred) or stargazer
• Include all standard elements: coefficients, SEs, significance stars, N, R-squared
• Export as .tex for LaTeX inclusion and .html for quick viewing

Figures:

• Use ggplot2 with ggthemes::theme_stata()
• Set bg = "transparent" for Beamer compatibility
• Include proper axis labels (sentence case, units)
• Export with explicit dimensions: ggsave(width = X, height = Y)
• Save as both .pdf and .png

Phase 5: Save and Review

1. fst::write_fst() for tabular data, qs2::qs_save() for model objects
2. Create output/ subdirectories as needed with dir.create(..., recursive = TRUE)
3. Run the r-reviewer agent on the generated script:

Delegate to the r-reviewer agent:
"Review the script at code/[script_name].R"

4. Address any Critical or High issues from the review.

Script Structure

Follow this template:

r
1# ============================================================
2# [Descriptive Title]
3# Author: [from project context]
4# Purpose: [What this script does]
5# Inputs: [Data files]
6# Outputs: [Figures, tables, cached data]
7# ============================================================
8
9# 0. Setup ----
10library(data.table)
11library(fixest)
12library(modelsummary)
13library(ggplot2)
14library(ggthemes)
15library(here)
16library(fst)
17library(qs2)
18
19dir.create(here("output", "analysis"), recursive = TRUE, showWarnings = FALSE)
20
21# 1. Data Loading ----
22# dt <- fread(here("data", "raw", "dataset.csv"), encoding = "UTF-8")
23
24# 2. Exploratory Analysis ----
25# [Summary stats, diagnostic plots]
26
27# 3. Main Analysis ----
28set.seed(888)
29# [Regressions, estimation — seed before any randomness step]
30
31# 4. Tables and Figures ----
32# [Publication-ready output with theme_stata()]
33
34# 5. Export ----
35# fst::write_fst(dt_results, here("output", "analysis", "results.fst"))
36# qs2::qs_save(model_fit, here("output", "analysis", "model.qs2"))
37# ggsave(here("output", "analysis", "figure.pdf"), width = 12, height = 5, bg = "transparent")

Important

• Reproduce, don't guess. If the user specifies a regression, run exactly that.
• Show your work. Print summary statistics before jumping to regression.
• Check for issues. Look for multicollinearity, outliers, perfect prediction.
• Use relative paths. All paths relative to repository root.
• Use here::here(). All paths via here() for cross-platform compatibility.
• No hardcoded values. Use variables for sample restrictions, date ranges, etc.