如何 plot seaborn histplot 的子图数量不均

Question

I currently have a list of 13 columns I am plotting a distribution of. I would like to create a series of subplots so that the plots take up less space but am having a difficult time doing so within a loop.

Sample DataFrame:

import pandas as pd
import numpy as np

data = {'identifier': ['A', 'B', 'C', 'D'],
        'treatment': ['untreated', 'treated', 'untreated', 'treated'], 'treatment_timing': ['pre', 'pre', 'post', 'post'],
        'subject_A': [1.3, 0.0, 0.5, 1.6], 'subject_B': [2.0, 1.4, 0.0, 0.0], 'subject_C': [nan, 3.0, 2.0, 0.5],
        'subject_D': [np.nan, np.nan, 1.0, 1.6], 'subject_E': [0, 0, 0, 0], 'subject_F': [1.0, 1.0, 0.4, 0.5]}

df = pd.DataFrame(data)

  identifier  treatment treatment_timing  subject_A  subject_B  subject_C  subject_D  subject_E  subject_F
0          A  untreated              pre        1.3        2.0        NaN        NaN          0        1.0
1          B    treated              pre        0.0        1.4        3.0        NaN          0        1.0
2          C  untreated             post        0.5        0.0        2.0        1.0          0        0.4
3          D    treated             post        1.6        0.0        0.5        1.6          0        0.5

• It goes from subject_A to subject_M (13 total).
• What I am currently doing produces a 13 row, 1 column layout of 13 histograms. One for each subject, separated into 3 colors (pre, post and missing).

Here is what I currently have:

fig, axes = plt.subplots(3,5, sharex=True, figsize=(12,6))

for index, col in enumerate(COL_LIST):
    sns.histplot(
            df ,x=col, hue="time", multiple="dodge", bins=10, ax=axes[index,index % 3]
        ).set_title(col.replace("_", " "))
plt.tight_layout()

This definitely doesn't work. But I'm not sure if there's an easy way to define the axes without having to copy and paste this line 13 times and manually define the axes coordinates.

Using displot is somewhat troublesome because col_wrap errors out

ValueError: Number of rows must be a positive integer, not 0

(I believe this is due to presence of np.nan)

Answer 1

• It will be easier to use seaborn.displot , which is a FacetGrid , instead of seaborn.histplot .
  • Explore using row , col , and col_wrap to get the number of rows and column as desired.
• The subject_ columns must be stacked, to convert the dataframe to a tidy format, which can be done with .stack

import pandas as pd
import seaborn as sns

# convert the dataframe into a long form with stack
df_long = df.set_index(['identifier', 'treatment', 'treatment_timing']).stack().reset_index().rename(columns={'level_3': 'subject', 0: 'vals'})

# sort by subject
df_long = df_long.sort_values('subject').reset_index(drop=True)

# display(df_long.head())
  identifier  treatment treatment_timing    subject  vals
0          A  untreated              pre  subject_A   1.3
1          D    treated             post  subject_A   1.6
2          C  untreated             post  subject_A   0.5
3          B    treated              pre  subject_A   0.0
4          D    treated             post  subject_B   0.0

# plot with displot
sns.displot(data=df_long, row='subject', col='treatment', x='vals', hue='treatment_timing', bins=10)