postaplot

POSTA (Presentation Of Sincere Total Adquisitions) plots provide kernel-density aware scatter plots with a seaborn-like API.
Each point is jittered horizontally according to the estimated density at its vertical position, so the cloud of points approximates the underlying distribution when a lot of observations are availabe, and functios as a regular strip or swarm plot with few points, making it a good default plot type.

Installation

python -m pip install "https://github.com/mwappner/postaplot/archive/refs/heads/main.zip"

Requires Python ≥3.8, and depends on numpy, scipy, matplotlib, pandas. PyPI coming soon.

Quickstart

import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
import postaplot

df = pd.DataFrame({
    "group": np.repeat(["A","B","C"], 100),
    "y": np.r_[np.random.normal(0,1,100),
               np.random.normal(1,1,100),
               np.random.standard_t(4,100)]
})

fig, ax = plt.subplots()
postaplot.kde_scatter(df, x="group", y="y", ax=ax, seed=0)
plt.show()

---

Features

• KDE-based jitter → point cloud approximates density.
• Discrete or continuous hue
  • discrete → legend
  • continuous → colorbar
  • customizable order, mapping and normalization
• Dodge discrete subgroups into separate sub-clouds.
• Add boxplots with a simple toggle.
• Flexible marker styling:
  • aliases: ms, mfc, mec, mew work like in plot().
  • hollow=True for edge-only markers.
• Clouds are rasterized if they have too many points (configurable)
• User-facing API: kde_scatter
• Low-level engine: kde_scatter_engine (returns PathCollection).

---

Documentation

See the Cookbook for examples:

• Normal vs uniform jitter
• Jitter width control
• Discrete vs continuous hue and customization
• Hollow markers (edge-colored by hue)
• Dodging multiple groups
• Boxplots
• Colorbars and legends
• Custom edge vs face colors

---

License

MIT