Feature Request: Implement Bagplot (Bivariate Boxplot) for 2D and 3D Data Visualization

This proposal suggests adding support for Bagplots (also known as bivariate boxplots) as a new visualization type in Plotly.
Bagplots are a powerful extension of the traditional boxplot to two or more dimensions, providing a powerful vizualisation which shows statistical properties of 2-3 different variables.

While Plotly currently provides strong support for univariate distribution plots (box, violin, histogram), there is no direct tool to visualize bivariate distribution summaries in a similar “statistical boxplot” style.

The Bagplot, introduced by Rousseeuw, Ruts, and Tukey (1999), fills this gap by visualizing the data depth and outliers in 2D (and potentially 3D) space.
It provides an intuitive, robust alternative to kernel density estimates or convex hulls when exploring multivariate data.

Reference:
Rousseeuw, P. J., Ruts, I., & Tukey, J. W. (1999). The Bagplot: A Bivariate Boxplot. The American Statistician, 53(4), 382–387.
ResearchGate link

Figure 1: Example of a 2D Bagplot showing data depth and outliers (adapted from Rousseeuw et al., 1999).

Proposed Functionality

Using the plotly api:

Express:

import plotly.express as px

# 2D Bagplot
fig2d = px.bagplot(df, x="feature1", y="feature2", depth_method="tukey", show_outliers=True)
fig2d.show()

# 3D Bagplot
fig3d = px.bagplot_3d(df, x="feature1", y="feature2", z="feature3",
                      depth_method="tukey", show_outliers=True)
fig3d.show()

Graph Objects:

import plotly.graph_objects as go

fig = go.Figure(data=go.Bagplot(
    x=df["feature1"],
    y=df["feature2"],
    show_outliers=True,
    depth_method="tukey"
))
fig.show()

fig = go.Figure(data=go.Bagplot3D(
    x=df["x"], y=df["y"], z=df["z"],
    show_outliers=True,
    depth_method="tukey"
))
fig.show()

Aucun fichier d’implémentation ni test n’est nommé. Commencez par examiner les API Python Express et graph_objects de Plotly ainsi que l’article Bagplot cité, puis clarifiez si le périmètre inclut la 2D, la 3D ou les deux ; le travail est considéré comme terminé lorsqu’une API convenue et un comportement de visualisation correspondant aux fonctionnalités de data-depth et de détection des valeurs aberrantes demandées sont définis.