Search code examples
pythonseabornbar-chartcatplot

How to set bar widths equal in catplot with horizontal bars

I have a data that I want to present as horizontal bar plots in rows and columns. I'm trying to do this with Python seaborn, which facet facility suits this purpose well.

The set of y-axis variables is different on the row dimension, and I don't want to present the empty items in facets. The problem is that the facets get the same heights while bar widths adjust. I wonder whether I could do this the opposite way so that all the bar widths would be equal in all facets, but the facet heights would adjust.

Here is an example:

import plotly.express as px
import seaborn as sns
df = px.data.tips()
g = sns.catplot(x="total_bill", y="day", col="sex", row="time", hue="smoker", data=df, 
                kind="bar", sharey=False, margin_titles=True)

And this is what the figure looks like (with seaborn version 0.11.2):

enter image description here

Solution
  • how to change the gridspec

    To get the output you want you can set the FacetGrid's gridspec according to the number of categories:

    df = px.data.tips()
df = df[df["day"] != "Thur"]
ratios = df.groupby('time')['day'].nunique().values
g = sns.catplot(x="total_bill", y="day", col="sex", row="time",
                data=df, kind="bar", sharey=False,
                legend=False, margin_titles=True,
                facet_kws={'gridspec_kws':{'height_ratios': ratios}})

    output:

    enter image description here

    previous answer: reproducibility

    If I reproduce your code with seaborn 0.11.0 I get the following output. Due to the Categorical values of day, the common days are broadcasted to all plots and the columns have the same width.

    >>> df = sns.load_dataset("tips")
>>> df.dtypes

total_bill     float64
tip            float64
sex           category
smoker        category
day           category
time          category
size             int64
dtype: object

    enter image description here

    update: plotly express dataset
    >>> import plotly.express as px
>>> df2 = px.data.tips()
>>> df2.dtypes
total_bill    float64
tip           float64
sex            object
smoker         object
day            object
time           object
size            int64
dtype: object