Managing subplots¶

One of the main feature of the library is that the slicer dimension is shared across subplot grids. This is convenient when drawing different graphic visualization along common data facets.

Statsplotly provides utility methods to further tweak subplots grid.

import plotly.express as px
from plotly.subplots import make_subplots

import statsplotly

Drawing subplots¶

All statsplotly plotting methods can be targeted to a plotly.graph_objects.Figure object created with plotly’s make_subplots function.

By default, subplot axes are independent :

df = (
    px.data.experiment().melt(id_vars=["gender", "group"], var_name="experiment").set_index("group")
)

fig = make_subplots(cols=2, rows=2, column_widths=[0.9, 0.1])

for row, group in enumerate(df.index.unique(), 1):
    fig = statsplotly.distplot(
        fig=fig, row=row, col=1, data=df.loc[group], y="value", slicer="gender", barmode="overlay"
    )

    fig = statsplotly.catplot(
        fig=fig,
        row=row,
        col=2,
        data=df.loc[group],
        plot_type="stripplot",
        x="group",
        slicer="gender",
        y="value",
    )

fig.layout.height = 800
fig.show()

Setting common axis limits¶

The set_common_axis_limit method of the SubplotGridFormatter class controls axis behaviour across the subplot grid.

Here we set common limits and link yaxes across the grid rows (while preserving the different limits of each row) :

from statsplotly.utils import SubplotGridFormatter

SubplotGridFormatter(fig=fig).set_common_axis_limit(shared_grid_axis="rows", link_axes=True)
fig.show()

Tidying subplots¶

SubplotGridFormatter also exposes the tidy_subplots utility method for flexible annotation of a subplot grid. The method automatically set a shared title for linked axes :

SubplotGridFormatter(fig=fig).tidy_subplots(
    title="Effect of treatment",
    row_titles=["Control", "Treatment"],
    col_titles=["Distribution", "Raw Data"],
)
fig.show()

Combining slicer and coloraxis across subplots¶

By chaining these methods, it is possible to combine and flexibly link subplots depicting different facets of a dataset.

Here we link y axes across rows, x axes across all subplots, and set a shared coloraxis across rows :

fig = make_subplots(cols=2, rows=2, row_heights=[0.2, 0.8])
for col, group in enumerate(df.index.unique(), 1):
    fig = statsplotly.distplot(
        fig=fig,
        col=col,
        row=2,
        data=df.loc[group],
        x="value",
        slicer="gender",
        hist=False,
        kde=True,
    )

    fig = statsplotly.catplot(
        fig=fig,
        col=col,
        row=1,
        plot_type="stripplot",
        orientation="horizontal",
        data=df.loc[group],
        color="experiment",
        x="value",
        slicer="gender",
        y="group",
        color_palette=["#d4f542", "#4275f5"],
    )

fig.layout.height = 800

SubplotGridFormatter(fig=fig).set_common_axis_limit(
    shared_grid_axis="rows", plot_axis="yaxis", link_axes=True
).set_common_axis_limit(
    shared_grid_axis="all", plot_axis="xaxis", link_axes=True
).set_common_coloraxis(
    shared_grid_axis="rows"
).tidy_subplots(
    title="Sex difference in response to treatment",
    col_titles=["control", "treatment"],
    row_titles=["raw data", "distribution"],
)

fig.show()

This last plot provides the following insights :

both sexes appear to respond to treatment positively, but the response effect size looks higher in males because of a lower baseline.
an experimental effect on the response to treatment is apparent in the scatter plot. This has to be taken into account in follow up statistical analyses.

Full details of the API : utils.