Social network outputs from nodeXL and Ucinet

Jul 29, 2013 at 5:53 PM
I am currently using several SNA tools to analyze the data.
However, the outputs from nodeXL are slightly different from the ones from Ucinet and R.
This slight difference in outputs determines the significance in statistical results of my study.
I understand basic theoretical formula for each social network measure such as degree, betweenness, etc.
I assume that you used them for developing nodeXL.
I would like to know what causes this issue and any solution for this.
Thank you!

Jul 30, 2013 at 6:04 PM
Edited Jul 30, 2013 at 6:05 PM

NodeXL calculates a few graph metrics itself, but it uses a software library called SNAP from Jure Leskovec's group at Stanford ( ) to calculate most of them. SNAP is a high-performance graph analysis package used by a variety of applications.

I haven't performed a side-by-side comparison of NodeXL's graph metrics with those calculated by UCINET or R, so I can't comment on the detailed differences between the results they produce. However, here are a few possible explanations that come to mind:
  1. Rounding errors in the various calculations.
  2. Rounding in Excel's displayed values. You may see the value "0.042" shown for a Closeness Centrality in the NodeXL workbook, for example, but the actual value in Excel's formula bar is 0.041667.
  3. Results may be normalized in NodeXL but not in UCINET and R, or vice-versa.
  4. The different packages may use different algorithms.
-- Tony
Aug 1, 2013 at 9:02 PM
Edited Aug 1, 2013 at 9:03 PM
Thank you for the reply!

-- Jinie