Parallel axis dot plot with vertical histograms for paired samples

Question

I've got some paired continuous data, and would like to partition the data into 'bins' or categories of equal size; and then create a similar plot to the attached image using Python & MatplotLib. The plot combines a parallel line plot to show the differences between 'before' and 'after' values of each data point with two centrally aligned vertical histograms to show the distribution of the data in each group:

The above example was apparently done using XLSTAT-PRO, but I've been unable to find any similar examples on the web which can do this using either MatplotLib or Pandas.

I was going to try to write a Python/MatplotLib routine myself, but wondered if anyone had already done something similar?

I'd be very grateful for any links/help and advice. Thanks in anticipation.

Answer 1

You can start reading this tutorial.

Here is something you can adapt for your problem:

import matplotlib.pyplot as plt
import numpy as np

# your input data:
befores = np.random.rand(10)
afters = np.random.rand(10)

# plotting the points
plt.scatter(np.zeros(len(befores)), befores)
plt.scatter(np.ones(len(afters)), afters)

# plotting the lines
for i in range(len(befores)):
    plt.plot( [0,1], [befores[i], afters[i]], c='k')

plt.xticks([0,1], ['before', 'after'])

plt.show()

Output:

Answer 2

Unable to find any Matplotlib or Pandas examples to do the above, so had a go at writing something myself. First attempt attached with dummy data and explanatory comments. I'm not a professional programmer, so apologies for the inelegant style (I'm always grateful for the opportunity to learn any tips & constructive feedback from the experts on this forum, and to share my code for anyone to improve); but this example works, and hopefully it, or parts of it, will be useful to someone somewhere...

import numpy as np
import matplotlib.pyplot as plt  
#--------------------------------
def points(x,y,n): # Plot n points symmetrically aligned about axes
    dx=0.03  # define distance between individual dots    
    m = 1-(n%2) # ensure symmetrical alignment for odd or even number of dots
    while(m<n):
        plt.scatter(x+(dx*m),y,color = 'k', marker = 'o', s=50, zorder=1)
        plt.scatter(x-(dx*m),y,color = 'k', marker = 'o', s=50, zorder=1)
        m+=2 
    return   
#--------------------------------   
def histogram(b): # count number of data points in each bin
    for col in range(0,2):
        count = np.unique(b[:,col], return_counts=True)  
        for n in range(0,np.size(count[col])):
            points(col,count[0][n], count[1][n])
    return
#-------------------------------        
def partition(a,bins): # partition continuous data into equal sized bins for plotting     
    lo = np.min(a)
    hi = np.max(a)
    rng = hi-lo
    step = rng/float(bins-1)

    for col in range (0,2):
        for row in range (0,int(np.size(a,axis=0))):
            for n in range (0,bins):
             if (a[row,col] <= (lo + (step / 2) + n * step)):
                 b[row,col] = (lo + (n * step))
                 break
    return(b)    
#--------------------------------    
def lines(b): # draw 'before' and 'after' lines between paired data points + median line
    for row in range (0,int(np.size(a,axis=0))):
        plt.plot([0,1],[b[row,0], b[row,1]], c='k',zorder=0, lw=1, alpha=0.3)
        plt.plot ([0,1],[np.median(b[:,0]),np.median(b[:,1])],c='r',zorder=2, lw=2, alpha=1)
    return
#================================    
# MAIN   
# Dummy paired continuous data (...or import from spreadsheet as a numpy array)\;    
a = np.array([
 [1.62,1.53,1.42,1.39,1.11,1.20,0.99,0.88,0.60,0.65,0.52,0.49,0.43,0.41,0.31], # before
[0.8,0.7,0.52,0.61,0.44,0.43,0.49,0.33,0.44,0.39,0.20,0.29,0.37,0.19,0.00] ]) # after

bins = 10 # choose total number of bins to categorise data into
ax=plt.axes()

a = a.transpose()
b=a # make a copy of the input data matrix to write categorised data to
b = partition(a,bins) # partition continuous data into bins
lines(b) # draw lines between mid points of each bin and draw median line
histogram(b) # draw histograms centered at mid points of each bin

# Make general tweaks to plot appearance here:
plt.xticks([0,1], ['OUT', 'IN'], fontsize=14)
plt.ylabel('stimulation threshold (mA)',fontsize=13)
plt.text(0.8,1.3,'All patients',fontsize=13)
ax.patch.set_facecolor('white') # set background colour for plot area (default = white)
ax.spines['top'].set_visible(False)   # remove default upper axis
ax.spines['right'].set_visible(False) # remove default right axis
plt.tick_params(axis='both',which='both',direction = 'out',top='off', right = 'off',labeltop='off') # remove tick marks from top & right axes
plt.xlim(-0.6,1.6)
plt.ylim(0,1.7)

plt.show()