Seaborn Jointplot add colors for each class

Question

I want to plot the correlation plot of 2 variables using seaborn jointplot. I have tried a lot of different things but I am not able to add colors to the points according to class.

Here is my code:

import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
sns.set()

X = np.array([5.2945 , 3.6013 , 3.9675 , 5.1602 , 4.1903 , 4.4995 , 4.5234 ,
              4.6618 , 0.76131, 0.42036, 0.71092, 0.60899, 0.66451, 0.55388,
              0.63863, 0.62504, 0.     , 0.     , 0.49364, 0.44828, 0.43066,
              0.57368, 0.     , 0.     , 0.64824, 0.65166, 0.64968, 0.     ,
              0.     , 0.52522, 0.58259, 1.1309 , 0.     , 0.     , 1.0514 ,
              0.7519 , 0.78745, 0.94873, 1.0169 , 0.     , 0.     , 1.0416 ,
              0.     , 0.     , 0.93648, 0.92801, 0.     , 0.     , 0.89594,
              0.     , 0.80455, 1.0103 ])

y = np.array([ 93, 115, 107, 115, 110, 107, 102, 113,  95, 101, 116,  74, 102,
               102,  78,  85, 108, 110, 109,  80,  91,  88,  99, 110, 108,  96,
               105,  93, 107,  98,  88,  75, 106,  92,  82,  84,  84,  92, 115,
               107,  97, 115,  85, 133, 100,  65,  96, 105, 112, 107, 107, 105])

ax = sns.jointplot(X, y, kind='reg' )
ax.set_axis_labels(xlabel='Brain scores', ylabel='Cognitive scores')
plt.tight_layout()
plt.show()

Now, I want to add colors for each point according to a class variable classes.

Answer 1

       label  Method 2  Method 1
0    Label 1  1.484914 -1.069439
1    Label 1  0.273158  1.139414
2    Label 1  1.089244  0.161752
3    Label 1  1.184306 -0.981758
4    Label 1  1.424435  0.300742
..       ...       ...       ...
111  Label 2 -0.201226  0.852319
112  Label 2  0.016911  0.985805
113  Label 2 -0.263775  0.248942
114  Label 2  3.283341 -1.247014
115  Label 2  0.325648  1.793694
[116 rows x 3 columns]

sns.jointplot(data=data, x="Method 1, y="Method 2", "hue="label", palette={
    'Label 1': '#d7191c',
    'Label 2': '#2b83ba'
})

Use joint_kws={"alpha": 0.5} to set transparency.

Example plot:

Answer 2

To build off Ernest's answer:

After you set scatter = False in sns.jointplot build the scatterplot using sns.scatterplot with the hue = classes argument equal to the categorical variable array. I find it cleanest to merge your data into a pandas dataframe with the columns x, y and classes and use this as the data for the scatterplot, but you don't have to do it this way...

classes = np.array([1., 1., 1., 1., 1., 1., 1., 1., 2., 2., 2., 2., 2., 2., 2.,
                    2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 
                    2., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 
                    3., 3., 3., 3., 3., 3., 3.])

# make them look a little more 'categorical'
classes = classes.astype('int')

x = np.array([5.2945 , 3.6013 , 3.9675 , 5.1602 , 4.1903 , 4.4995 , 4.5234 ,
              4.6618 , 0.76131, 0.42036, 0.71092, 0.60899, 0.66451, 0.55388,
              0.63863, 0.62504, 0.     , 0.     , 0.49364, 0.44828, 0.43066,
              0.57368, 0.     , 0.     , 0.64824, 0.65166, 0.64968, 0.     ,
              0.     , 0.52522, 0.58259, 1.1309 , 0.     , 0.     , 1.0514 ,
              0.7519 , 0.78745, 0.94873, 1.0169 , 0.     , 0.     , 1.0416 ,
              0.     , 0.     , 0.93648, 0.92801, 0.     , 0.     , 0.89594,
              0.     , 0.80455, 1.0103 ])

y = np.array([ 93, 115, 107, 115, 110, 107, 102, 113,  95, 101, 116,  74, 102,
               102,  78,  85, 108, 110, 109,  80,  91,  88,  99, 110, 108,  96,
               105,  93, 107,  98,  88,  75, 106,  92,  82,  84,  84,  92, 115,
               107,  97, 115,  85, 133, 100,  65,  96, 105, 112, 107, 107, 105])

sns.jointplot(x, y, kind='reg', scatter = False )
sns.scatterplot(x, y, hue=classes)

Answer 3

I managed to find a solution that is exactly what I need. Thank to @ImportanceOfBeingErnest that gave me the idea to let the regplot only draw the regression line.

Solution:

import pandas as pd

classes = np.array([1., 1., 1., 1., 1., 1., 1., 1., 2., 2., 2., 2., 2., 2., 2.,
                    2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 2., 
                    2., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 3., 
                    3., 3., 3., 3., 3., 3., 3.])

df = pd.DataFrame(map(list, zip(*[X.T, y.ravel().T])))
df = df.reset_index()
df['index'] = classes[:]

g = sns.jointplot(X, y, kind='reg', scatter = False )
for i, subdata in df.groupby("index"):
    sns.kdeplot(subdata.iloc[:,1], ax=g.ax_marg_x, legend=False)
    sns.kdeplot(subdata.iloc[:,2], ax=g.ax_marg_y, vertical=True, legend=False)
    g.ax_joint.plot(subdata.iloc[:,1], subdata.iloc[:,2], "o", ms = 8)
plt.tight_layout()
plt.show()

Answer 4

The obvious solution is to let the regplot only draw the regression line, but not the points and add those via a usual scatter plot, which has the color c argument.

g = sns.jointplot(X, y, kind='reg', scatter = False )
g.ax_joint.scatter(X,y, c=classes)