Tukukkk

I am writing code to plot cross correlations of every time-series in my data against all others, with two for-loops to index the row and column position respectively (column loop nested within the row loop).

Currently only the final axes (i.e. bottom right corner) of the figure is displaying any data, and each iteration of the loop appears to be plotting on this axes. I am wondering if I have made any obvious mistakes with the order of commands in the nested for loops, or if I am misinterpreting the input arguments to matplotlib functions like subplots....

The code is as below:

fig, axes = plt.subplots(nrows=data_num, ncols=data_num, sharex=True, sharey=True)



for n in range(data_num):  #row index

    for p in range(data_num):  # column index

        x = data_df.iloc[:,n]  #get data for ROI according to row index

        print(x.head())

        x = x.values

        y = data_df.iloc[:,p]  #get data for ROI according to column index

        print(y.head())

        y = y.values

        axes[n,p] = plt.xcorr(x,y,normed=True)     #axes [row,column] = cross correlation plot of above data

        print(f'plotting at index [ {n} , {p}]')

It's a, perhaps unfortunate, way that pyplot and matplotlib works: you have to create the plots on the respective axes, not assign the result from a pyplot.xcorr call to an axes. Thus: axis[n,p].xcorr(...). So the interface is suddenly somewhat more object-oriented than the usual direct pyplot calls.

All the plots ends up in just the last figure, because you are calling

plt.xcorr(x,y,normed=True)

It doesn't matter if you then assign the return value to the axes array elements, which you shouldn't, as that destroys the original axes array.
plt.xcorr will then plot all the data in the same plot on top of each other, because pyplot generally acts on the currently active axes, which is the last one created via plt.subplots().

That's for an explanation. Here's an example solution (with random data and a simple scatter plot):

import numpy as np

import matplotlib.pyplot as plt



data_num = 3



x = np.random.uniform(1, 10, size=(data_num, data_num, 20))

y = np.random.uniform(5, 20, size=(data_num, data_num, 20))



fig, axes = plt.subplots(nrows=data_num, ncols=data_num, sharex=True, sharey=True)

for n in range(data_num):  #row index

    for p in range(data_num):  # column index

        # Call `scatter` or any plot function on the 

        # respective `axes` object itself

        axes[n,p].scatter(x[n,p], y[n,p])

        print(f'plotting at index [ {n} , {p}]')

plt.savefig('figure.png')

and figure.png looks like (sorry, no colour or symbol variation, just bare bones scatter plots):