Correlation Matrix

Jun 4, 2014 at 3:27 PM
I want to create a network from a correlation matrix. Here is a short version:
A   b   C   D   E   F
A 1.00 -0.02 0.11 0.16 -0.04 -0.04
B -0.02 1.00 -0.05 -0.03 0.00 -0.02
C 0.11 -0.05 1.00 0.51 -0.01 -0.02
D 0.16 -0.03 0.51 1.00 -0.01 -0.01
E -0.04 0.00 -0.01 -0.01 1.00 0.84
F -0.04 -0.02 -0.02 -0.01 0.84 1.00

Higher values represent a closer relation of the words A and B whereas a negativ value describes a negative association.
Can NodeXL process those values to a correct network or do I need to change them, for example rank them from 1-10 with 10 representing the greatest connection?
Does the calumn 'edge weight' expect values within a certain range?

I hope you can see my problem and help me or forward me to a similar discussion.


sorry about the poor formatting
Jun 5, 2014 at 1:28 AM
Edited Jun 5, 2014 at 6:49 AM
Hello, Sonja:

Oh, don't worry about formatting. Your question is perfectly understandable.

There are no restrictions on edge weight values, except that they have to be numeric. They can be positive, negative, or zero; whole numbers or fractional.

However, NodeXL makes only limited use of the Edge Weight column that it creates on the Edges worksheet when you use the Import from Open Matrix Workbook feature.* Please see this post for more information about that:

-- Tony

* NodeXL, Data, Import, From Open Matrix Workbook in the Excel ribbon.
Jun 5, 2014 at 7:53 AM
Thanks for the quick reply.

So am I right in assuming that NodeXL only uses edge weights within the Fruchterman-Reingold layout algorithm?
Are there any further possibilities to influence the location of my vertices in Fruchterman-Reingold layout?
Jun 5, 2014 at 5:07 PM
Not on a per-vertex basis, no. There are a few layout settings that affect the overall algorithm (they're at NodeXL, Graph, Layout, Layout Options, Fruchterman-Reingold layout), but that's probably not what you're seeking.

-- Tony
Jun 10, 2014 at 8:09 AM
Hi Tony,
thanks for your answer.
Okay so I guess I will have to work with the Fruchterman-Reingold layout algorithm.
One last question:
What exactly does the 'lay out again' button do? How does it change the network? And how can I find the right number of klicks on the button to get an optimal result?
Jun 10, 2014 at 4:42 PM
The Fruchterman-Reingold layout algorithm uses the current vertex locations as a starting point. "Lay Out Again" runs the algorithm again. Because Fruchterman-Reingold mimics attractive and repulsive forces between particles, the net result of clicking "Lay Out Again" is often that the vertices that are repelling each other end up farther apart.

I don't know if there is a correct number of clicks, or if there is even such a thing as an optimal result. Perhaps others reading this can offer advice.

-- Tony