Errorbar variable marker size

Question

I want to plot some data x and y in which I need the marker size to depend on a third array z. I could plot them separately (i.e., scatter x and y with size = z, and errorbar without marker, fmc = 'none') and this solves it. The problem is that I need the legend to show the errorbar AND the dot, together:

and not

Code is here with some made-up data:

import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(1,10,100)
y = 2*x
yerr = np.random(0.5,1.0,100)
z = np.random(1,10,100)

fig, ax = plt.subplots()

plt.scatter(x, y, s=z, facecolors='', edgecolors='red', label='Scatter') 
ax.errorbar(x, y, yerr=yerr, xerr=0, fmt='none', mfc='o', color='red', capthick=1, label='Error bar')

plt.legend()

plt.show()

which produces the legend I want to avoid:

In errorbar the argumentmarkersizedoes not accept arrays asscatter` does.

Answer 1

The idea is usually to use a proxy to put into the legend. So while the errorbar in the plot may have no marker, the one in the legend has a marker set.

import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(1,10,11)
y = 2*x
yerr = np.random.rand(11)*5
z = np.random.rand(11)*2+5

fig, ax = plt.subplots()

sc = ax.scatter(x, y, s=z**2, facecolors='', edgecolors='red') 
errb = ax.errorbar(x, y, yerr=yerr, xerr=0, fmt='none', 
                   color='red', capthick=1, label="errorbar")

proxy = ax.errorbar([], [], yerr=[], xerr=[], marker='o', mfc="none", mec="red", 
                    color='red', capthick=1, label="errorbar")
ax.legend(handles=[proxy], labels=["errorbar"])
plt.show()