brycewang-stanford/Auto-Empirical-Research-Skills / plugins/empirical-analysis-r/skills/pipeline
Full-empirical-analysis-skill-R
VerifiedClassical end-to-end empirical analysis workflow in the modern tidyverse + econometrics R ecosystem — dplyr + tidyr + haven + fixest + sandwich + lmtest + clubSandwich + AER + ivreg + did + bacondecomp + HonestDiD + eventstudyr + rdrobust + rddensity + Synth + gsynth + synthdid + MatchIt + WeightIt + cobalt + ebal + grf + DoubleML + mediation + marginaleffects + modelsummary + kableExtra + gt + ggplot2 + ggpubr + cowplot + binsreg. **Defaults to economics empirical-paper style** (AER / QJE / AEJ) — every run produces a publication-ready output set with a multi-column regression table (M1→M6 progressive controls/FE) as the centerpiece, plus Table 1 (descriptives), mechanism / heterogeneity / robustness tables, and event-study + coefficient + trend figures. Covers the full 8-step R pipeline an applied economist runs on every paper — (1) data import & cleaning (read_dta/read_csv, naniar, janitor, validate-merges), (2) variable construction (mutate/across/winsorize/group_by + lag/lead with dplyr), (3) descriptive statistics & Table 1 (gtsummary, modelsummary::datasummary, tableone), (4) classical diagnostic tests (shapiro/jarque.bera.test/bptest/dwtest/bgtest/vif/adf.test/kpss.test/Hausman), (5) baseline modeling (fixest::feols, ivreg, did::att_gt, eventstudyr, sun_ab, did_imputation, synthdid, rdrobust, MatchIt, WeightIt, grf::causal_forest, DoubleML, mediation), (6) robustness battery (modelsummary stack, clubSandwich CRSE, fwildclusterboot, ri2, robomit Oster, bacondecomp, HonestDiD), (7) further analysis (interactions + marginaleffects, mediation::mediate, gsem via lavaan, dose-response splines, grf CATE), (8) publication-ready tables & figures (modelsummary, kableExtra, gt, stargazer, texreg, flextable to LaTeX/Word/HTML; ggplot2 + ggpubr + cowplot + binsreg + iplot for figures). **Also covers two parallel domain modes that share the same 8-step scaffolding** — **Mode A — Epidemiology / public health** (target-trial emulation, IPTW + g-formula + TMLE doubly-robust
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 121.7 KB |
| references/01-data-cleaning.md | markdown | 12.2 KB |
| references/02-data-transformation.md | markdown | 11.2 KB |
| references/03-descriptive-stats.md | markdown | 12.8 KB |
| references/04-statistical-tests.md | markdown | 10.4 KB |
| references/05-modeling.md | markdown | 20.5 KB |
| references/06-robustness.md | markdown | 14.6 KB |
| references/07-further-analysis.md | markdown | 13.2 KB |
| references/08-tables-plots.md | markdown | 15.0 KB |
references/03-descriptive-stats.md
View on GitHub# Step 3 — Descriptive Statistics & Table 1 in R (Deep Reference)
Goal: before any regression, produce the full descriptive set a referee expects — sample size, central tendency, dispersion, comparison across treatment, balance with SMDs, correlation matrix with significance stars, and the distribution / time-trend plots that motivate identification.
R's modern table ecosystem (`gtsummary`, `modelsummary::datasummary*`, `tableone`) makes this almost a one-liner.
## Contents
1. Full-sample summary (`gtsummary::tbl_summary`, `datasummary_skim`, `psych::describe`)
2. Stratified Table 1 (treated vs control) with SMDs
3. Weighted descriptives (`survey`, `gtsummary` weights)
4. Correlation matrix with significance + heatmap
5. Distribution plots (`ggplot2`, `ggdist`, density / ECDF / Q-Q)
6. Group comparisons (box / violin / strip / sina via `ggdist`)
7. Time-trend (DID motivation) plot
8. Panel coverage diagnostics
9. Binscatter via `binsreg`
10. Export recipes (LaTeX, Word, HTML, Excel, PNG/PDF)
---
## 1. Full-sample summary
### `gtsummary` — modern Table 1 standard
```r
library(gtsummary)
library(dplyr)
df %>%
select(log_wage, age, edu, tenure, training, female) %>%
tbl_summary(
type = list(all_continuous() ~ "continuous2"),
statistic = list(
all_continuous() ~ c("{N_nonmiss}", "{mean} ({sd})",
"{min} – {median} – {max}"),
all_categorical() ~ "{n} ({p}%)"
),
digits = list(all_continuous() ~ 3),
label = list(log_wage ~ "Log monthly wage",
age ~ "Age (years)",
edu ~ "Years of education",
tenure ~ "Tenure (years)",
training ~ "Received training",
female ~ "Female")
) %>%
bold_labels() %>%
add_n()
```
### `modelsummary::datasummary*` — fast & format-flexible
```r
library(modelsummary)
# One-shot summary
df %>%
select(log_wage, age, edu, tenure, training) %>%
datasummary_skim(output = "tables/table1_full.tex")
# Custom layout
datasummary(
log_wage + age + edu + tenure ~ N + Mean + SD + Min + Median + Max,
data = df,
output = "tables/table1_full.html"
)
```
### `psych::describe` — quick numeric summary
```r
library(psych)
df %>% select(log_wage, age, edu, tenure) %>% describe()
# Outputs: vars, n, mean, sd, median, trimmed, mad, min, max, range, skew, kurtosis, se
```
For categoricals, frequencies:
```r
df %>%
count(industry, sort = TRUE) %>%
mutate(pct = n / sum(n) * 100)
```
---
## 2. Stratified Table 1 (treated vs control + SMDs)
The single most-read table in any empirical paper.
### `gtsummary` route — most polished
```r
df %>%
select(log_wage, age, edu, tenure, female, training) %>%
tbl_summary(by = training, missing = "ifany",
statistic = all_continuous() ~ "{mean} ({sd})") %>%
add_p() %>% # Wilcoxon / chi-sq automatically
add_difference() %>% # group difference
add_overall() %>%
modify_header(label = "**Variable**") %>%
modify_caption("**Table 1.** Sample characteristics by treatment status") %>%
bold_labels() %>%
as_kable_extra(format = "latex", booktabs = TRUE) %>%
kableExtra::save_kable("tables/table1_balance.tex")
```
### `modelsummary::datasummary_balance` — fastest LaTeX path
```r
datasummary_balance(
~ training,
data = df %>% select(training, age, edu, tenure, female),
output = "tables/table1_balance.tex",
fmt = 3,
dinm = TRUE, # difference-in-means column
dinm_statistic = "p.value" # add p-value
)
```
### `tableone` route — gives clean SMD column
```r
library(tableone)
vars <- c("log_wage", "age", "edu", "tenure", "female")
t1 <- CreateTableOne(vars = vars, strata = "training", data = df,
factorVars = "female", test = TRUE)
print(t1, smd = TRUE, showAllLevels = TRUE,
catDigits = 1, contDigits = 3, missing = TRUE,
printToggle = FALSE) %>%
kableExtra::kable("latex", booktabs = TRUE) %>%
kableExtra::save_kable("tables/table1_tableone.tex")
```
**Interpretation rules**:
| |SMD| | Reading |
|------|---------|
| < 0.10 | well-balanced |
| 0.10 – 0.25 | modest imbalance — control in regression |
| > 0.25 | severe — consider matching / weighting |
---
## 3. Weighted descriptives
```r
library(survey)
library(gtsummary)
# Declare survey design
svy_design <- svydesign(ids = ~psu, strata = ~stratum, weights = ~svy_wt, data = df)
# Weighted Table 1
svy_design %>%
tbl_svysummary(
by = training,
include = c(log_wage, age, edu, tenure, female)
) %>%
add_p() %>%
add_overall()
# Weighted summary stat (one-off)
weighted.mean(df$wage, df$svy_wt, na.rm = TRUE)
Hmisc::wtd.mean(df$wage, df$svy_wt)
Hmisc::wtd.var(df$wage, df$svy_wt)
Hmisc::wtd.quantile(df$wage, weights = df$svy_wt, probs = c(0.25, 0.5, 0.75))
```
---
## 4. Correlation matrix with significance
```r
library(psych)
library(corrplot)
cols <- c("log_wage","age","edu","tenure","training")
# corr.test gives r, p, n with adjustment for multiple testing
ct <- corr.test(df %>% select(all_of(cols)), method = "pearson", adjust = "BH")
ct$r # correlation matrix
ct$p # p-values (upper tri = adjusted, lower = unadjusted)
# Heatmap with significance markers (insig = "blank" hides non-sig)
corrplot(ct$r, method = "color", type = "upper",
p.mat = ct$p, sig.level = 0.05, insig = "blank",
addCoef.col = "black", number.cex = 0.7,
tl.col = "black", tl.srt = 45,
col = colorRampPalette(c("#B2182B","white","#2166AC"))(200),
title = "", mar = c(0,0,1,0))
# Save:
pdf("figures/corr_heatmap.pdf", width = 6, height = 5)
corrplot(...)
dev.off()
# Or via ggcorrplot (ggplot-native)
library(ggcorrplot)
ggcorrplot(ct$r, p.mat = ct$p, type = "upper",
hc.order = TRUE, lab = TRUE, lab_size = 3,
outline.color = "white",
colors = c("#B2182B", "white", "#2166AC"),
insig = "blank", sig.level = 0.05) +
labs(title = "Correlation matrix (asterisks = p < 0.05)")
ggsave("figures/corr_ggcorr.pdf", width = 6, height = 5)
# Spearman (rank-based)
ct_sp <- corr.test(df %>% select(all_of(cols)), method = "spearman")
```
For LaTeX export of correlations:
```r
library(modelsummary)
datasummary_correlation(df %>% select(all_of(cols)),
output = "tables/corr.tex")
```
---
## 5. Distribution plots
```r
library(ggplot2)
# Histogram + density
p_hist <- ggplot(df, aes(wage)) +
geom_histogram(bins = 50, fill = "navy", alpha = 0.7) +
labs(title = "(a) Histogram", x = "Wage", y = "Count") +
theme_classic()
# KDE by group
p_kde <- ggplot(df, aes(log_wage, fill = factor(training))) +
geom_density(alpha = 0.5) +
scale_fill_manual(values = c("0" = "darkred", "1" = "navy"),
labels = c("Control","Treated"), name = "") +
labs(title = "(b) KDE by treatment", x = "Log wage", y = "Density") +
theme_classic() + theme(legend.position = "bottom")
# ECDF
p_ecdf <- ggplot(df, aes(log_wage, color = factor(training))) +
stat_ecdf(geom = "step", linewidth = 0.7) +
scale_color_manual(values = c("0" = "darkred", "1" = "navy"),
labels = c("Control","Treated"), name = "") +
labs(title = "(c) ECDF", x = "Log wage", y = "Cumulative share") +
theme_classic() + theme(legend.position = "bottom")
# Q-Q vs Normal
p_qq <- ggplot(df, aes(sample = log_wage)) +
stat_qq(alpha = 0.3) + stat_qq_line(color = "red") +
labs(title = "(d) Q-Q vs Normal", x = "Theoretical", y = "Sample") +
theme_classic()
library(cowplot)
cowplot::plot_grid(p_hist, p_kde, p_ecdf, p_qq, ncol = 2) +
ggsave("figures/distributions.pdf", width = 10, height = 7)
# Two-sample KS
ks.test(df$log_wage[df$training == 1],
df$log_wage[df$training == 0])
```
---
## 6. Group comparisons
```r
library(ggplot2)
library(ggdist) # half-eye / sina / dotplot — modern alternative to box+violin
ggplot(df, aes(factor(training), log_wage)) +
ggdist::stat_halfeye(adjust = 0.5, width = 0.5,
.width = 0, justification = -0.2) +
geom_boxplot(width = 0.15, outlier.shape = NA, alpha = 0.5) +
ggdist::stat_dots(side = "left", justification = 1.2,
binwidth = 0.025, alpha = 0.3) +
scale_x_discrete(labels = c("0" = "Control", "1" = "Treated")) +
labs(x = "", y = "Log wage", title = "Distribution comparison")
ggsave("figures/group_compare.pdf", width = 6, height = 4)
```
---
## 7. Time-trend (DID motivation) plot
```r
trend <- df %>%
group_by(year, training) %>%
summarise(mean_log_wage = mean(log_wage, na.rm = TRUE),
n = n(), .groups = "drop")
ggplot(trend, aes(year, mean_log_wage, color = factor(training))) +
geom_line(linewidth = 1) +
geom_point(size = 2.5) +
geom_vline(xintercept = policy_year, linetype = "dashed", color = "gray40") +
annotate("text", x = policy_year, y = max(trend$mean_log_wage),
label = "Policy", hjust = -0.1, color = "gray40") +
scale_color_manual(values = c("0" = "darkred", "1" = "navy"),
labels = c("Control","Treated"), name = "") +
labs(x = "Year", y = "Mean log wage",
title = "Pre/post trends by treatment status") +
theme_classic() + theme(legend.position = "bottom")
ggsave("figures/trend_did.pdf", width = 7, height = 4)
# Difference (treated − control) — pre-period should hug zero
trend_diff <- trend %>%
pivot_wider(id_cols = year, names_from = training, values_from = mean_log_wage,
names_prefix = "g") %>%
mutate(diff = g1 - g0)
ggplot(trend_diff, aes(year, diff)) +
geom_line(linewidth = 1, color = "navy") +
geom_point(size = 2.5, color = "navy") +
geom_hline(yintercept = 0, linetype = "dashed") +
geom_vline(xintercept = policy_year, linetype = "dashed", color = "gray40") +
labs(y = "Δ log wage (treated − control)", x = "Year")
ggsave("figures/did_diff.pdf", width = 7, height = 4)
```
---
## 8. Panel coverage diagnostics
```r
# Units per year
df %>%
count(year, name = "n_units") %>%
ggplot(aes(year, n_units)) +
geom_col(fill = "navy", alpha = 0.8) +
labs(x = "Year", y = "# unique observations") +
theme_classic()
ggsave("figures/panel_coverage.pdf", width = 7, height = 4)
# Years per unit
df %>%
count(worker_id) %>%
ggplot(aes(n)) + geom_histogram(bins = 30, fill = "navy", alpha = 0.8) +
labs(x = "# years observed per worker", y = "Count") +
theme_classic()
ggsave("figures/years_per_unit.pdf", width = 6, height = 4)
# Cohort sizes (staggered DID)
df %>%
filter(!is.na(first_treat)) %>%
distinct(worker_id, first_treat) %>%
count(first_treat) %>%
ggplot(aes(first_treat, n)) + geom_col(fill = "navy", alpha = 0.8) +
labs(x = "First treatment year", y = "# units") + theme_classic()
ggsave("figures/cohort_sizes.pdf", width = 6, height = 4)
# Observation-matrix heatmap
df %>%
mutate(obs = 1) %>%
ggplot(aes(year, factor(worker_id), fill = obs)) +
geom_tile() +
scale_fill_gradient(low = "white", high = "black", guide = "none") +
labs(x = "Year", y = "Worker", title = "Observed (black) vs missing (white)") +
theme_classic() + theme(axis.text.y = element_blank())
ggsave("figures/obs_heatmap.pdf", width = 8, height = 5)
```
---
## 9. Binscatter
```r
library(binsreg)
# Default — residualized + CI band
bs <- binsreg(y = df$log_wage, x = df$tenure,
w = df %>% select(age, edu, female),
nbins = 20, polyreg = 1, ci = c(2,2))
# bs$bins_plot is a ggplot object — save it:
bs$bins_plot + labs(x = "Tenure", y = "Residualized log wage")
ggsave("figures/binscatter_resid.pdf", width = 7, height = 4)
# By group
binsreg(y = df$log_wage, x = df$tenure, by = df$training,
w = df %>% select(age, edu),
nbins = 20)$bins_plot
```
---
## 10. Export recipes
```r
# LaTeX
datasummary_skim(df %>% select(...), output = "tables/x.tex")
gtsummary_obj %>%
as_kable_extra(format = "latex", booktabs = TRUE) %>%
kableExtra::save_kable("tables/x.tex")
# Word
gtsummary_obj %>%
as_flex_table() %>%
flextable::save_as_docx(path = "tables/x.docx")
modelsummary(..., output = "tables/x.docx")
# HTML
gtsummary_obj %>% as_gt() %>% gt::gtsave("tables/x.html")
# Excel
df %>% writexl::write_xlsx("tables/x.xlsx")
# Figures (always both PDF for LaTeX + PNG for Word/web)
ggsave("figures/x.pdf", width = 7, height = 4)
ggsave("figures/x.png", width = 7, height = 4, dpi = 300)
```
---
## Standard Step 3 deliverable
For every paper, produce these 6 artifacts in Step 3:
1. `tables/table1_full.tex` — full-sample summary
2. `tables/table1_balance.tex` — treated vs. control + SMDs
3. `figures/corr_heatmap.pdf` — correlation heatmap with stars
4. `figures/distributions.pdf` — 2×2 panel (hist + KDE + ECDF + Q-Q)
5. `figures/trend_did.pdf` — DID motivation plot
6. `figures/panel_coverage.pdf` (or `cohort_sizes.pdf` for staggered designs)
Once all 6 exist and look right, you can move to Step 4 with full confidence about the sample.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the work | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixes | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation code | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |