python scatter plot area size proportional axis length

问题

I'm getting quite desperate about this, I couldn't find anything on the www so far.

Here's the situation:

I am working with Python.
I have 3 arrays: the x-coordinates, the y-coordinates and the radius.
I want to create a scatter plot with the given x- and y-coordinates.

So far, everything works how I want it to. Here is what's bothering me:

The circle size of each point in the scatter plot should be defined by the radius array.
The the values of the coordinates and the radius are in same units. More explicitly: Let's assume I have a point at (1, 1) with radius 0.5 assigned. Then I want to get a circle in the plot centered at (1, 1) and with the border going throught the points (1.5, 1), (1, 1.5), (0.5, 1) and (1, 0.5)

What I am struggling with is to find out the ratio of plot points to the length in an axis. I need to work with points because as far as I can see, the circle size of the scatter plot is given in point values. So if let's say my axis goes from 0 to 10, I need to know how many points there are in between in the plot.

Can anybody help me? Or is there another way of doing this? Thanks in advance.

回答1:

I'm jumping in from your other stackoverflow question. I think the approach you presented as an answer to the present question won't work exactly as you want for the following reasons:

First, the size of the markers is in points, not in pixels. In typography, the point is the smallest unit of measure and correspond in matplotlib to a fixed length of 1/72 inch. In contrast, the size of a pixel will vary following the figure dpi and size.
Second, the size of the markers in plt.scatter are related to the diameter of the circles, not the radius.

So the size in points of each marker should be calculated as:

size_in_points = (2 * radius_in_pixels / fig_dpi * 72 points/inch)**2

Moreover, as shown in the MWE below, it is possible to calculate the size of the marker radius in pixels directly with matplotlib transformations, without having to generate an empty figure beforehand:

import numpy as np
import matplotlib.pyplot as plt

plt.close('all')

# Generate some data :
N = 25
x = np.random.rand(N) + 0.5
y = np.random.rand(N) + 0.5
r = np.random.rand(N)/10

# Plot the data :
fig = plt.figure(facecolor='white', figsize=(7, 7))
ax = fig.add_subplot(111, aspect='equal')
ax.grid(True)
scat = ax.scatter(x, y, s=0, alpha=0.5, clip_on=False)
ax.axis([0, 2, 0, 2])

# Draw figure :
fig.canvas.draw()

# Calculate radius in pixels :
rr_pix = (ax.transData.transform(np.vstack([r, r]).T) -
          ax.transData.transform(np.vstack([np.zeros(N), np.zeros(N)]).T))
rpix, _ = rr_pix.T

# Calculate and update size in points:
size_pt = (2*rpix/fig.dpi*72)**2
scat.set_sizes(size_pt)

# Save and show figure:
fig.savefig('scatter_size_axes.png')
plt.show()

A point at (1, 1) with radius 0.5 assigned will result in a circle in the plot centered at (1, 1) and with the border going throught the points (1.5, 1), (1, 1.5), (0.5, 1) and (1, 0.5):

来源：https://stackoverflow.com/questions/36458458/python-scatter-plot-area-size-proportional-axis-length

标签

python

axis

scatter-plot

area

scatter