在 Altair 的分组和堆叠条形图中为每个组使用单独的比例

Question

I'd like to use separate scales for each group in a grouped and stacked barplot using Altair in Python.

So for example instead of the following

I'd like something similar to the following. In this (Gimp-)edited picture I have the same scale for all the 4 groups A,B,C and D. But In my actual data, the orders of magnitude are different from a group to another. So each Group should have a different scale. Any ideas on how to do that?

Here is an Minimum Example from HERE

import pandas as pd
import numpy as np
import altair as alt

df1=pd.DataFrame(10*np.random.rand(4,3),index=["A","B","C","D"],columns=["I","J","K"])
df2=pd.DataFrame(10*np.random.rand(4,3),index=["A","B","C","D"],columns=["I","J","K"])
df3=pd.DataFrame(10*np.random.rand(4,3),index=["A","B","C","D"],columns=["I","J","K"])

def prep_df(df, name):
    df = df.stack().reset_index()
    df.columns = ['c1', 'c2', 'values']
    df['DF'] = name
    return df

df1 = prep_df(df1, 'DF1')
df2 = prep_df(df2, 'DF2')
df3 = prep_df(df3, 'DF3')

df = pd.concat([df1, df2, df3])

chart = alt.Chart(df).mark_bar().encode(
    x=alt.X('c2:N', title=None),
    y=alt.Y('sum(values):Q', axis=alt.Axis(grid=False, title=None)),
    column=alt.Column('c1:N', title=None),
    color=alt.Color('DF:N', scale=alt.Scale(range=['#96ceb4', '#ffcc5c','#ff6f69']))
).configure_view(
    strokeOpacity=0    
)

chart.save("Power.svg")

Answer 1

You can have independent axes for the charts by adding

resolve_scale(y='independent')

Note that, by itself, this lets the y-domain limits for each facet adjust to the subset of the data within each facet; you can make them match by explicitly specifying domain limits.

Put together, it looks like this:

alt.Chart(df).mark_bar().encode(
    x=alt.X('c2:N', title=None),
    y=alt.Y('sum(values):Q', axis=alt.Axis(grid=False, title=None), scale=alt.Scale(domain=[0, 25])),
    column=alt.Column('c1:N', title=None),
    color=alt.Color('DF:N', scale=alt.Scale(range=['#96ceb4', '#ffcc5c','#ff6f69']))
).configure_view(
    strokeOpacity=0    
).resolve_scale(
    y='independent'
)