tidyecon

broom + modelsummary for Python econometrics.

tidyecon closes the last gap between Python and Stata/R for applied microeconomics:

• tidy() — unified coefficient table from any supported model (mirrors R broom::tidy)
• glance() — unified model statistics (mirrors R broom::glance)
• modelsummary() — publication-ready regression tables → HTML, LaTeX, Word (mirrors R modelsummary)
• tidyecon-validate — CLI numerical benchmark suite against hardcoded R results

Supports statsmodels, pyfixest, and linearmodels via dispatch. Zero modifications to upstream model objects.

---

Installation

uv add tidyecon                    # core (statsmodels)
uv add "tidyecon[all]"             # + pyfixest + linearmodels
uv add "tidyecon[pyfixest]"        # individual backend
uv add "tidyecon[linearmodels]"

Requires Python ≥ 3.11.

---

Quick start

import statsmodels.api as sm
import pandas as pd
import tidyecon as te

X      = sm.add_constant(df[["income", "education"]])
fit    = sm.OLS(df["wages"], X).fit()
fit_hc = sm.OLS(df["wages"], X).fit(cov_type="HC1")

te.tidy(fit)     # DataFrame: term, estimate, std_error, statistic, p_value, conf_low, conf_high
te.glance(fit)   # DataFrame: nobs, r_squared, adj_r_squared, rmse, f_statistic, ...

te.modelsummary(
    {"OLS": fit, "OLS (HC1)": fit_hc},
    coef_map={"const": "Intercept", "income": "Income", "educ": "Education"},
    notes=["HC1 robust SEs in column 2."],
    output="table.html",   # or .tex or .docx
)

---

API

tidy(model, conf_level=0.95)

Returns a DataFrame with columns: term · estimate · std_error · statistic · p_value · conf_low · conf_high

Supported backends:

Backend	Types
statsmodels	Any ResultsWrapper — OLS, WLS, GLS, Logit, Probit, Poisson, …
pyfixest	Feols, Fepois
linearmodels	PanelOLS, PooledOLS, RandomEffects, BetweenOLS, FirstDifferenceOLS, IV2SLS, IVGMM

glance(model)

Returns a single-row DataFrame with model statistics: nobs · r_squared · adj_r_squared · rmse · f_statistic · p_value_f · df_model · df_residual · estimator · fixed_effects · vcov_type

modelsummary(models, *, stars, statistic, fmt, coef_map, coef_omit, gof_map, title, notes, output)

Parameter	Default	Options
stars	True	True · False · {"***": 0.01, ...}
statistic	"se"	"se" · "tstat" · "pvalue" · "confint"
output	"html"	"html" · "latex" · "docx" · any *.html/*.tex/*.docx path

# Rename and reorder coefficients (drops terms not in map)
te.modelsummary(models, coef_map={"wt": "Weight", "hp": "Horsepower"})

# Drop intercept
te.modelsummary(models, coef_omit=["const"])

# t-statistics instead of SEs
te.modelsummary(models, statistic="tstat")

# Write directly to file
te.modelsummary(models, output="results/table.docx")

---

Validation

tidyecon-validate            # run all R benchmarks
tidyecon-validate --verbose  # show every check

Compares tidy() output against hardcoded R 4.4.1 values (OLS + Logit on mtcars). Add new fixtures in src/tidyecon/validate/fixtures.py.

---

Development

git clone https://github.com/yourorg/tidyecon && cd tidyecon
uv sync --extra dev
uv pip install -e .  # Install in editable mode for development
uv run pytest
uv run ruff check src tests
uv run pyright src
uv run tidyecon-validate --verbose

Important setup notes:

• uv sync --extra dev installs all development dependencies including pytest, statsmodels, etc.
• uv pip install -e . makes the tidyecon module importable during development
• Tests require statsmodels and other optional dependencies to be installed

See TODO.md for the roadmap and CONTRIBUTING.md for contribution guidelines.

---

License

MIT — see LICENSE.

Inspired by: R broom · R modelsummary · pyfixest