matplotlib graphs in SPSS

Question

Is it possible to use any other graphing library in SPSS that the built in? I just discovered the python extensions that makes SPSS great.

import matplotlib.pyplot as plt
from numpy.random import rand


fig, ax = plt.subplots()
for color in ['red', 'green', 'blue']:
    n = 750
    x, y = rand(2, n)
    scale = 200.0 * rand(n)
    ax.scatter(x, y, c=color, s=scale, label=color,
               alpha=0.3, edgecolors='none')

ax.legend()
ax.grid(True)

plt.show() 

This will create a simple scatter plot and it works fine in any IDE, but when trying to use that code in SPSS BEGIN PROGRAM END PROGRAM i get the following error:

RuntimeError: Python is not installed as a framework. The Mac OS X backend will not be able to function correctly if Python is not installed as a framework. See the Python documentation for more information on installing Python as a framework on Mac OS X.
 Please either reinstall Python as a framework, or try one of the other backends. If you are Working with Matplotlib in a virtual enviroment see 'Working with Matplotlib in Virtual environments' in the Matplotlib FAQ

Maybe I am asking too much out of the python extension in SPSS but it would be nice to use another graph library than the one they have built in.

Answer 1

You can use just about any Python code or library, but you need to do some configuring. When the Python support (Python Essentials) is installed, it installs a private, unregistered Python system in order not to conflict with any other Python that might be installed. So if you try to add other libraries, the installer doesn't know what to do or installs them somewhere that the Statistics installation won't know about.

The easiest way to get around this is to install another, standard Python installation (version 2.7 in recent versions or 3.4 with version 24 if you need Python 3). Then go to Edit > Options > Files and point to that distribution. You will need to restart Statistics for that to take effect. I use the Anaconda distribution, which includes a lot of other goodies.

I'm not a Mac user, but something like this should work.

Answer 2

Adding more information as another answer as the comment field is too limited.

I do not get the framework error on Windows, which is a different issue, I think. But running your code inside Statistics works - sort of. Instead of placing the image in the Viewer, it pops up in its own window (which may be buried behind another window).

So Statistics doesn't know about it and patiently waits for the program to complete, which doesn't happen until you dismiss that window (which does have a normal frame not shown in the graphic).

To make this work, you would need to direct the matplotlib code to write the image to a file somewhere and then use SpssClient apis to insert that image in the Viewer. See CreateImageChartItem Method (Python) in the Python programmability help. Alternatively, if you can direct matplotlib to write the image to the standard output stream, Statistics might be able to capture it directly in the Viewer.

I usually do programmability images with R code (even though Python is a way better language(!)), where this all works seamlessly. Or I use Python code to write Statistics graphics commands and GPL to have the Statistics engine, which is pretty powerful, do the charting.

Another thing you mind find helpful if you are into Python, is that you can run Python in external mode, where you start with Python code running from your IDE or a Python command line and then have it invoke Statistics by running

import spss

This has great advantages for developing and debugging Python code, but you can't use the SpssClient module methods directly. I ran your code from my IDE (Wing Professional), and the image window popped right up. And, of course, you can use the spss module and related apis in external mode to communicate with Statistics and control it.