plotting correlation matrix using python

Question

I want to plot the correlation matrix using python. I have tried with the following script

  corr_matrix=np.corrcoef(vector)
  imshow(corr_matrix, interpolation='bilinear')
  colorbar()
  show()

The dimension of the matrix is 2500X2500. The above code produces a matrix of full of dots. But I want smooth surface. How do I get that.

Best Sudipta

Aman · Accepted Answer · 2013-02-06 00:07:03Z

5

What do you mean by "smooth surface" and why do you want to visualize your correlation matrix that way?

Here are two useful examples for visualizing [correlation] matrices. Both contain an explanation as well as example code for matplotlib.

Square grid pseudocolor plot http://glowingpython.blogspot.com/2012/10/visualizing-correlation-matrices.html
Hinton Diagram http://www.scipy.org/Cookbook/Matplotlib/HintonDiagrams

Update: To supplement my comment, here's a pseudocolor visualization of a 1000x1000 correlation matrix, which didn't encounter memory issues on my humble laptop:

enter image description here

Note that although row 20 is correlated to other variables and row 40 is correlated to row 80, in the style of the GlowingPython example, yet this information is obscured by the sheer size of the matrix.

edited Feb 6, 2013 at 0:07

answered Feb 2, 2013 at 22:37

Aman

47.4k7 gold badges38 silver badges37 bronze badges

Sign up to request clarification or add additional context in comments.

9 Comments

user1964587 Over a year ago

the pcolor() method is out of memory for my matrix. It works for small matrix (say 10X10) but for the large matrix it is not working. Is there any other method which work similar to the pcolor()?

Aman Over a year ago

I'm not sure looking at a pcolor chart of a matrix that large (2500x2500) will tell you anything useful. Having said that, how large do you get before the memory error? Perhaps consider plotting a quarter of the matrix at a time?

Aman Over a year ago

See my update which contains a pic of a 1000x1000 corr matrix.

Aman Over a year ago

Same script as in the first link (GlowingPython), I just changed the size.

Joe Kington Over a year ago

@user1964587 - On a side note, use pcolormesh (or just imshow as you're already doing, but with interpolation='nearest') instead of pcolor for large arrays. (pcolormesh is limited to rectangular cells, whereas pcolor isn't, thus the speedup.) The advantage (or disadvantage) to using pcolormesh or pcolor over imshow when displaying regular grids is mostly that the former produce vector output. At any rate, pcolormesh should solve your problems with the large matrix.

|

sriisking · Accepted Answer · 2014-04-30 10:12:45Z

0

You can sort the columns based on the values obtained in the correlation matrix.

answered Apr 30, 2014 at 10:12

sriisking

11 bronze badge

Collectives™ on Stack Overflow

plotting correlation matrix using python

2 Answers 2

9 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

9 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related