How to set the line style for each kdeplot in a jointgrid

Question

I am using seaborn to create a kdeplot with marginal distribution as described in this answer. I adapted the code a little to give me this:

import matplotlib.pyplot as plt
import seaborn as sns

iris = sns.load_dataset("iris")
setosa = iris.loc[iris.species == "setosa"]
virginica = iris.loc[iris.species == "virginica"]

g = sns.JointGrid(x="sepal_width", y="petal_length", data=iris)
sns.kdeplot(setosa.sepal_width, setosa.sepal_length, cmap="Reds",
        shade=False, shade_lowest=False, ax=g.ax_joint)
sns.kdeplot(virginica.sepal_width, virginica.sepal_length, cmap="Blues",
        shade=False, shade_lowest=False, ax=g.ax_joint)
sns.distplot(setosa.sepal_width, kde=True, hist=False, color="r", ax=g.ax_marg_x)
sns.distplot(virginica.sepal_width, kde=True, hist=False, color="b", ax=g.ax_marg_x)
sns.distplot(setosa.sepal_length, kde=True, hist=False, color="r", ax=g.ax_marg_y, vertical=True)
sns.distplot(virginica.sepal_length, kde=True, hist=False, color="b", ax=g.ax_marg_y, vertical=True)
plt.show()

This is impossible to print in black and white. How can I get seaborn to print the kdeplot and distplot lines in a specifically styled (dotted / dashed / ...) way to make them distinguishable when printed in black and white?

The related questions deal with other types of plots which appear to support this, but this does not seem to be supported by kdeplot and distplot.

Answer 1

Marginals

To show the lines of a kde plot with a different linestyle you use the linestyle argument, which is passed on to matplotlib's plot function.

sns.kdeplot(setosa.sepal_width, color="r", ax=g.ax_marg_x, linestyle="--")

To provide this argument to the kde plot which is produced via distplot you may use the kde_kws argument.

sns.distplot(..., kde_kws={"linestyle":"--"})

However, there does not seem to be any reason to use a distplot here.

Joint KDE

For the 2D case, the linestyle argument has no effect. A 2D kdeplot is a contour plot. Hence you need to use the contour's linestyles (not the s) argument.

sns.kdeplot(, linestyles="--")

Complete code

import matplotlib.pyplot as plt
import seaborn as sns

iris = sns.load_dataset("iris")
setosa = iris.loc[iris.species == "setosa"]
virginica = iris.loc[iris.species == "virginica"]

g = sns.JointGrid(x="sepal_width", y="petal_length", data=iris)

sns.kdeplot(setosa.sepal_width, setosa.sepal_length, cmap="Reds",
        shade=False, shade_lowest=False, ax=g.ax_joint, linestyles="--")
sns.kdeplot(virginica.sepal_width, virginica.sepal_length, cmap="Blues",
        shade=False, shade_lowest=False, ax=g.ax_joint, linestyles=":")

sns.kdeplot(setosa.sepal_width, color="r", legend=False,  
             ax=g.ax_marg_x, linestyle="--")
sns.kdeplot(virginica.sepal_width, color="b", legend=False, 
             ax=g.ax_marg_x, linestyle=":")
sns.kdeplot(setosa.sepal_length, color="r", legend=False, 
             ax=g.ax_marg_y, vertical=True, linestyle="--")
sns.kdeplot(virginica.sepal_length, color="b", legend=False, 
             ax=g.ax_marg_y, vertical=True, linestyle=":")
plt.show()

Answer 2

Any keywords that sns.kdeplot does not recognise are passed to plt.contour() or plt.contourf(). In your case it is contourf, so you can pass the keyword linestyles (note the plural). sns.distplot has a keyword that is called kde_kws, which accepts a dictionary of keywords that are passed to plt.plot. In this case you can use ls or linestyle (note the singular). Below is a complete example:

import matplotlib.pyplot as plt
import seaborn as sns

iris = sns.load_dataset("iris")
setosa = iris.loc[iris.species == "setosa"]
virginica = iris.loc[iris.species == "virginica"]

g = sns.JointGrid(x="sepal_width", y="petal_length", data=iris)
sns.kdeplot(
    setosa.sepal_width, setosa.sepal_length, cmap="Greys",
    shade=False, shade_lowest=False, ax=g.ax_joint,
    linestyles='--'
)
sns.kdeplot(
    virginica.sepal_width, virginica.sepal_length, cmap="Greys",
    shade=False, shade_lowest=False, ax=g.ax_joint,
    linestyles=':'
)
sns.distplot(
    setosa.sepal_width, kde=True, hist=False, color="k",
    kde_kws=dict(ls ='--'), ax=g.ax_marg_x
)
sns.distplot(
    virginica.sepal_width, kde=True, hist=False, color="k",
    kde_kws=dict(ls=':'), ax=g.ax_marg_x
)
sns.distplot(
    setosa.sepal_length, kde=True, hist=False, color="k",
    kde_kws=dict(ls ='--'), ax=g.ax_marg_y, vertical=True
)
sns.distplot(
    virginica.sepal_length, kde=True, hist=False, color="k",
    kde_kws=dict(ls=':'), ax=g.ax_marg_y, vertical=True
)
plt.show()

And the result looks like this: