问题
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.
回答1:
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)
回答2:
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()
回答3:
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)
来源：https://stackoverflow.com/questions/51210955/seaborn-jointplot-add-colors-for-each-class