Facetted grouped boxplot

In the seaborn plotting library exists catplot() that allows facetted grouped boxplots:

import pandas as pd

## the following code needs the "xlrd" library installed for reading Excel files.
## alternatively, you can use the attached data.csv which contains the same data using:
# dataProp = pd.read_csv('https://github.com/altair-viz/altair/files/4565572/data.zip')

link = 'https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/812408/Gas_and_Electric_Consumption_Multiple_Attributes_Table_2019.xlsx'
data = pd.read_excel(link,'2017 Consumption')
dataGas = data.loc[data['Gas present']=='Yes',]
dataProp = dataGas.loc[dataGas['Electricity type']=='Standard',]

import seaborn as sns

sns.catplot(x='Number of bedrooms',
            y='Median - Gas', 
            col='Region', 
            hue='Property age',
            kind="box",
            data=dataProp,
            col_wrap=4,
            hue_order=['Pre 1919','1919-44','1945-64','1965-1982','1983-92','1993-99'])

I tried to replicate this using altair, but the boxplots get stacked on themselfs:

import altair as alt

alt.Chart(dataProp).mark_boxplot().encode(
    x='Number of bedrooms:O',
    y='Median - Gas',
    color=alt.Color('Property age:O',
                    sort=['Pre 1919','1919-44','1945-64','1965-1982','1983-92','1993-99']),
    facet=alt.Facet('Region:N', 
                    columns=4),
).interactive()

I tried to combine facet=... and column=... parameters in encode() and tried my hands on the repeat() method without success. Is there a way to show the grouped boxplots next to each other like in seaborn’s catplot()?

Thank you for you help.

[jakevdp]

No, there’s no way to do this yet in Altair/Vega-Lite (the way to make grouped boxplots is to use facet, and vega-lite does not support facets of facets).

A future vega-lite release will hopefully include an offset encoding which will make this type of chart possible.

[joelostblom]

@VietTralala Sorry I should have mentioned that this has been added in the development version on GitHub, but it has not yet been released. As you noticed, you can install directly from GitHub to get access to the X/YOffset encodings. I am not sure why it is not working for you, but you should not have to do anything special with the import, so try importing as usual import altair as alt. It is important that you close down all notebooks (or clear their output) and the reopen jupyterlab. If there is a plot open that has been created by an with an old version of vegalite, that can create issues.