moderndive (Python)¶

The Python companion package for ModernDive: Statistical Inference via Data Science — a faithful port of the R moderndive and infer packages to a modern Python data-science stack (polars, plotly, plotnine, statsmodels).

If you teach or learn statistical inference the tidy way — specify → hypothesize → generate → calculate → visualize — this package gives you the same grammar in Python, plus regression helpers and the book’s datasets.

Why moderndive?¶

A tidy inference grammar. Bootstrap confidence intervals and permutation/simulation hypothesis tests read like sentences, mirroring R infer.
Regression helpers that return tidy tables: get_regression_table, get_regression_points, get_regression_summaries, get_correlation.
Dual-engine plots. Every plot takes engine="plotly" (default, interactive) or engine="plotnine" (grammar-of-graphics) — same code, your choice of output.
58 bundled datasets via load_*() loaders returning polars DataFrames.
polars-first, works with pandas too.

Installation¶

pip install moderndive

Optional extra for saving plotly figures as static images (PNG/SVG):

pip install "moderndive[image]"   # adds kaleido

30-second example¶

Are tracks more likely to be popular in metal than in deep house? Compute the observed difference in “popular” rates, then shuffle the genre labels 1000 times to build a null distribution and read off a p-value.

import moderndive as md
from moderndive import specify, observe, get_p_value, visualize, shade_p_value

spotify = md.load_spotify_metal_deephouse()

# Observed difference in "popular" proportions (metal − deep-house)
obs = observe(
    spotify, formula="popular_or_not ~ track_genre", success="popular",
    stat="diff in props", order=("metal", "deep-house"),
)
obs

ObservedStatistic(stat='diff in props', value=0.034)

# Null distribution under "genre doesn't matter" (permutation)
null = (
    spotify.specify(formula="popular_or_not ~ track_genre", success="popular")
    .hypothesize(null="independence")
    .generate(reps=1000, type="permute", seed=76)
    .calculate(stat="diff in props", order=("metal", "deep-house"))
)
get_p_value(null, obs_stat=obs, direction="right")

shape: (1, 1)

p_value
f64
0.075

# Visualize it (interactive plotly by default; engine="plotnine" also works)
visualize(null) + shade_p_value(obs_stat=obs, direction="right")

_images/740bf8671170f1b5361f3ad35be67c52f9ce7c9f19b3daff3e4152648e08beb7.png

Note

Plots throughout this documentation are rendered as static images. When you run the code yourself, the default engine="plotly" produces interactive figures (hover, zoom, pan); engine="plotnine" gives static grammar-of-graphics plots.

Where to next¶

New here? Start with Getting started.
Coming from R? See Coming from R for a function-by-function map.
Browse the task guides below, or jump to the API reference.

Get started

Guides

Reference

API reference