更改 seaborn 的因子图以包含点

Question

I have a pandas dataframe that looks like this:

      feat  roi       sbj         alpha test_type       acc
0     cnn2  LOC  Subject1  normal_space   imagery  0.260961
1     cnn2  LOC  Subject1           0.4   imagery  0.755594
2     cnn4  LOC  Subject1  normal_space   imagery  0.282238
3     cnn4  LOC  Subject1           0.4   imagery  0.726485
4     cnn6  LOC  Subject1  normal_space   imagery  0.087359
5     cnn6  LOC  Subject1           0.4   imagery  0.701167
6     cnn8  LOC  Subject1  normal_space   imagery  0.209444
7     cnn8  LOC  Subject1           0.4   imagery  0.612597
8    glove  LOC  Subject1  normal_space   imagery  0.263176
9    glove  LOC  Subject1           0.4   imagery  0.659182
10    cnn2  FFA  Subject1  normal_space   imagery  0.276830
11    cnn2  FFA  Subject1           0.4   imagery  0.761014
12    cnn4  FFA  Subject1  normal_space   imagery  0.288127
13    cnn4  FFA  Subject1           0.4   imagery  0.727325
14    cnn6  FFA  Subject1  normal_space   imagery  0.113507
15    cnn6  FFA  Subject1           0.4   imagery  0.732963
16    cnn8  FFA  Subject1  normal_space   imagery  0.264455
17    cnn8  FFA  Subject1           0.4   imagery  0.615467
18   glove  FFA  Subject1  normal_space   imagery  0.245950
19   glove  FFA  Subject1           0.4   imagery  0.640502
20    cnn2  PPA  Subject1  normal_space   imagery  0.344078
...

For plotting it, I wrote:

ax = sns.factorplot(x="feat", y="acc", col="roi", hue="alpha", alpha = 0.9, data=df_s_pt, kind="bar").set(title = "perception, scene wise correlation")

The result look like this:

I want to upgrade it so it can look like the one in this answer (so it has the dots of each subject (ie, Subject1, Subject2, ...))

Also, I want to control the color.

I could'nt use the code in that answer . How should I apply having dots/color change in factorplot?

Thanks in advance

Answer 1

Some remarks:

• sns.factorplot is a very old function. In the newer seaborn versions it has been replaced by sns.catplot . To take advantage of the hard work in correcting, improving and extending the library, it is highly recommended to upgrade to the latest version (0.12.2)
• Functions that create multiple subplots in one go, don't return an ax , but a grid of subplots (a FacetGrid ). It is extremely confusing storing the result of such a function in ax , as matplotlib's axes functions won't work on them.
• Calling set(title=...) on the FacetGrid changes the titles of the individual subplots. It therefore removes the title given by seaborn to indicate the feature used for each subplot ( 'roi' in the current example).
• To change the overall title, g.fig.suptitle(...) can be used. Some extra space needs to be provided, as that doesn't happen automatically.
• The latest seaborn versions have a function g.map_dataframe to apply a function to each subset used corresponding to its subplot.
• Colors can be controlled via the palette= parameter. Either individual colors, or a colormap can be chosen.
• To make sure the order is the same everywhere, it often helps to make the dataframe columns of type pd.Categorical .
• You might want to suppress the errorbars with sns.catplot(..., errorbar=None)

Here is an example starting from dummy test data.

from matplotlib import pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np

df = pd.DataFrame({'feat': np.random.choice(['cnn2', 'cnn4', 'cnn6', 'cnn8', 'glove'], 100),
                   'roi': np.random.choice(['LOC', 'FFA'], 100),
                   'alpha': np.random.choice(['normal_space', 0.4], 100),
                   'acc': 1 - np.random.rand(100) ** 2})
df['feat'] = pd.Categorical(df['feat'])
df['roi'] = pd.Categorical(df['roi'])
df['alpha'] = pd.Categorical(df['alpha'])

g = sns.catplot(x="feat", y="acc", col="roi", hue="alpha", palette=['crimson', 'limegreen'],
                alpha=0.9, data=df, kind="bar")
g.map_dataframe(sns.stripplot, x="feat", y="acc", hue="alpha", palette=['cornflowerblue', 'yellow'],
                edgecolor="black", linewidth=.75, dodge=True)
g.set(xlabel='')  # remove the xlabels if they are already clear from the xticks
g.fig.subplots_adjust(top=0.9)  # need extra space for the overall title
g.fig.suptitle("perception, scene wise correlation")
plt.show()