Overlay Contour Plots in Matplotlib

Question

I need to compare 2 dimensional distributions of 2 groups.

When I use matplotlib.pyplot.contourf and overlay the plots, the background color of each contour plot fills the entire plot space. Is there any way to make the lowest contour level transparent for each contour plot so that it's easier to see the center of each contour?

import numpy as np
import matplotlib.pyplot as plt
from matplotlib import cm 
import scipy.stats as st

def make_cloud(x, y, std, n=100):
    x = np.random.normal(x, std, n)
    y = np.random.normal(y, std, n)
    return np.array(zip(x, y))

def contour_cloud(x, y, cmap):
    xmin, xmax = -4, 4
    ymin, ymax = -4, 4

    xx, yy = np.mgrid[xmin:xmax:100j, ymin:ymax:100j]
    positions = np.vstack([xx.ravel(), yy.ravel()])
    values = np.vstack([x, y])
    kernel = st.gaussian_kde(values)
    f = np.reshape(kernel(positions).T, xx.shape)

    plt.contourf(xx, yy, f, cmap=cmap, alpha=0.5)


cloud1 = make_cloud(-1, 1, 1)
cloud2 = make_cloud(1, -1, 1)

plt.scatter(x=cloud1[:,0], y=cloud1[:,1])
plt.scatter(x=cloud2[:,0], y=cloud2[:,1], color='red')

fig = plt.gcf()
ax = plt.gca()

contour_cloud(x=cloud1[:, 0], y=cloud1[:, 1], cmap=cm.Blues)
contour_cloud(x=cloud2[:, 0], y=cloud2[:, 1], cmap=cm.Reds)

Answer 1

There are a few controls you will want to look at for contourf. You can manually change the different levels and you can change the color map over/under specifications. By default, the fill for areas under the lowest level (or above the max) seems to be transparent.

So, the easiest way to do what you want is to manually specify the levels and specify them such that there are points below the lowest level, but are not any points above the highest level.

If you replace:

plt.contourf(xx, yy, f, cmap=cmap, alpha=0.5)

with:

step = 0.02
m = np.amax(f)
levels = np.arange(0.0, m, step) + step
plt.contourf(xx, yy, f, levels, cmap=cmap, alpha=0.5)

produces an image like:

For more details on the behavior for values over/under the colormap values, see here.