Dual y-axis plot using seaborn objects (v0.12)

Question

I am trying to make a dual y-axis plot using the new seaborn interface (seaborn.objects, available in v0.12). However, I am having difficulties getting it to work.

My first try is this:

import seaborn.objects as so

df = pd.DataFrame(
    {"x": [1, 2, 3, 4, 5], "y1": [5, 2, 1, 6, 2], "y2": [100, 240, 130, 570, 120]}
)
fig, ax1 = plt.subplots(1, 1, figsize=(10, 5))
ax2 = ax1.twinx()
so.Plot(df, x="x").add(so.Bar(), y="y1", ax=ax1).add(so.Line(), y="y2", ax=ax2)

But this will create the seaborn plot with one y-axis and an empty dual-axis plot.

Answer 1

Here is the solution I finally came up with:

import seaborn.objects as so
import pandas as pd
import matplotlib.pyplot as plt


df = pd.DataFrame(
    {"x": [1, 2, 3, 4, 5], "y1": [5, 2, 1, 6, 2], "y2": [1000, 240, 1300, 570, 120]}
)
fig, ax1 = plt.subplots(1, 1, figsize=(6, 4))
ax2 = ax1.twinx()
ax1.tick_params(axis="x", labelrotation=45)
p1 = (
    so.Plot(df, x="x", y="y1")
    .add(so.Bar(width=0.7))
    .label(x="Month", y="Data 1", title="TITLE")
    .on(ax1)
    .plot()
)
p2 = (
    so.Plot(df, x="x", y="y2")
    .add(so.Line(color="orange", linewidth=3))
    .label(y="Data 2", title="TITLE")
    .scale(y=so.Continuous().label(like="{x:,.0f}"))
    .on(ax2)
    .plot()
)

The result is pretty much what I was looking for.

Now here is an interesting fact. The above image was generated with python 3.10.11 and seaborn 0.12.2.

When switching to python 3.11.4 and seaborn 0.13.2, I get the following (note the y-axis tick labels on the left:

This seems to be a bug in seaborn.

Answer 2

How can we refine this?

h= sns.load_dataset("healthexp")
fig, ax = plt.subplots(1, 1, figsize=(10, 5))
ax.clear()
plt.close()
ax2= ax.twinx()
p=(
so.Plot(h, x="Year")
)

p.add(so.Bar(),so.Agg('sum'), y="Spending_USD").on(ax).plot()
p=p.add(so.Line(),so.Agg(), y="Life_Expectancy").on(ax2)
p=p.label(
             x="", 
             y="",
             title="Spending",
        )
p

Even the output of original plot has messed up axis label.

df = pd.DataFrame(
    {"x": [1, 2, 3, 4, 5], "y1": [5, 2, 1, 6, 2], "y2": [100, 240, 130, 570, 120]}
)
fig, ax1 = plt.subplots(1, 1, figsize=(10, 5))
ax2 = ax1.twinx()
p = so.Plot(df, x="x")
p.add(so.Bar(), y="y1").on(ax1).plot()
p.add(so.Line(), y="y2").on(ax2).plot()

Answer 3

You'll want to call Plot.on to use pre-existing matplotlib axes:

import seaborn.objects as so

df = pd.DataFrame(
    {"x": [1, 2, 3, 4, 5], "y1": [5, 2, 1, 6, 2], "y2": [100, 240, 130, 570, 120]}
)
fig, ax1 = plt.subplots(1, 1, figsize=(10, 5))
ax2 = ax1.twinx()
p = so.Plot(df, x="x")
p.add(so.Bar(), y="y1").on(ax1).plot()
p.add(so.Line(), y="y2").on(ax2).plot()

Note that there will likely be a Plot.twin method to make this more natural, but it's not been implemented yet.