How to plot a scatter FacetGrid from a dataframe with multi-level columns

Question

data = {('Weight', 'Additive', 'Water'): {0: 3, 1: 3, 2: 3, 3: 3, 4: 3},
        ('Weight', 'Additive', 'Grass'): {0: 6.0, 1: 7.0, 2: 6.0, 3: 0, 4: 0},
        ('Weight', 'Filler', 'Flowers'): {0: 1, 1: 2, 2: 3, 3: 4, 4: 5},
        ('Color', '', ''): {0: 15, 1: 18, 2: 21, 3: 35, 4: 40}}
combo = pd.DataFrame(data)

    Weight               Color
  Additive        Filler      
     Water Grass Flowers      
0        3   6.0       1    15
1        3   7.0       2    18
2        3   6.0       3    21
3        3   0.0       4    35
4        3   0.0       5    40

I have a MultiIndex level table in a pandas dataframe called 'combo' and am trying to make a FacetGrid scatterplot. When I try with the code below I get the error: 'Length of order must be same as number of levels (3), got 2'

grid = sns.FacetGrid(combo, col = combo['Weight'])
grid.map(sns.scatterplot, combo['Weight'], combo['Color'])
plt.show()

user_7758804 · Answer

Tested in python 3.10, pandas 1.4.2, matplotlib 3.5.1, seaborn 0.11.2
It is not recommended to directly use sns.FacetGrid. For figure-level scatter and line plots, use sns.relplot.
See How to flatten a hierarchical index in columns for more options to flatten the multi-level columns.
The easiest way to plot the data is to flatten the column names, and then convert the dataframe to a long form with pandas.DataFrame.melt.

import seaborn as sns
import pandas as pd

# sample data
data = {('Weight', 'Additive', 'Water'): {0: 3, 1: 3, 2: 3, 3: 3, 4: 3},
        ('Weight', 'Additive', 'Grass'): {0: 6.0, 1: 7.0, 2: 6.0, 3: 0, 4: 0},
        ('Weight', 'Filler', 'Flowers'): {0: 1, 1: 2, 2: 3, 3: 4, 4: 5},
        ('Color', '', ''): {0: 15, 1: 18, 2: 21, 3: 35, 4: 40}}
df = pd.DataFrame(data)

# flatten the column titles and remove two spaces from the 'Color  ' column name
df.columns = df.columns.to_flat_index().str.join(' ').str.replace('  ', '')

# convert the df to a long format to use with seaborn
dfm = df.melt('Color', var_name='Substance', value_name='Weight')

# clean the column values in the Substance column
dfm.Substance = dfm.Substance.str.replace('Weight ', '').str.replace('_', ' ')

# display(dfm)
    Color       Substance  Weight
0      15  Additive Water     3.0
1      18  Additive Water     3.0
2      21  Additive Water     3.0
3      35  Additive Water     3.0
4      40  Additive Water     3.0
5      15  Additive Grass     6.0
6      18  Additive Grass     7.0
7      21  Additive Grass     6.0
8      35  Additive Grass     0.0
9      40  Additive Grass     0.0
10     15  Filler Flowers     1.0
11     18  Filler Flowers     2.0
12     21  Filler Flowers     3.0
13     35  Filler Flowers     4.0
14     40  Filler Flowers     5.0

plot using `hue`

g = sns.relplot(kind='scatter', data=dfm, x='Weight', y='Color', hue='Substance')

plot using `col`

g = sns.relplot(kind='scatter', data=dfm, x='Weight', y='Color', col='Substance')

How to plot a scatter FacetGrid from a dataframe with multi-level columns

1 Answers

plot using hue

plot using col

plot using `hue`

plot using `col`