Power changes in social networks

Nov 21, 2009 at 4:27 PM

Hello,

Have just recently mastered the basics of NodeXL (version 1.0.1.98) and want to start studying the relative changes in informal power within a social network as a result of various changes to it's structure and linkages. Can I measure the power of each vertex by adding together it's key metrics (ie. Degree, Betweenness and Closeness centrality and Eigenvector centrality) or is this too simplistic? Is there a better method possible with NodeXL?

Would appreciate your advice.

Wildboar100 

Coordinator
Nov 21, 2009 at 5:02 PM
Edited Nov 21, 2009 at 5:20 PM

Hello!  Thank you for your interest in NodeXL.

You are touching on issues of social network analysis theory and method that depend on details of the data and the research question.  You may want to review the International Network for Social Network Analysis site at http://www.insna.org - in particular, consider joining the SOCNET-L email list where a great group of people with deep SNA experience gather to discuss these kinds of questions.

My short answer is: no; I would not simply add those values together.  The "key metrics (ie. Degree, Betweenness and Closeness centrality and Eigenvector centrality)" all mean very different things.  

You may want to explore the ways these values change over time in the network.  Each metric has a description that could help you interpret what changes in those values mean in your dataset (see: http://en.wikipedia.org/wiki/Centrality and http://en.wikipedia.org/wiki/Social_network).

For example, increasing density for the whole graph (a metric that is reported on the Overall Metrics worksheet) suggests that more relationships are being created over the intervening time, while older relationships are being maintained.  If a particular node's eigenvector centrality increases significantly, they may have created new important relationships or their friends may have developed new connections.

NodeXL can implement almost any model you decide to create by building formulas that relate the node metrics and other attributes.  These compound metrics can be mapped to any display attribute using the Auto-fill columns feature.

Regards,

Marc

 

 

Nov 21, 2009 at 5:51 PM
Hello Marc,
Many thanks, will follow your advice.
Wildboar100
----- Original Message -----
From: [email removed]
To: [email removed]
Sent: Saturday, November 21, 2009 5:03 PM
Subject: Re: Power changes in social networks [NodeXL:75805]

From: marcsmith

Hello! Thank you for your interest in NodeXL.

You are touching on issues of social network analysis theory and method that depend on details of the data and the research question. You may want to review the International Network for Social Network Analysis site at http://www.insna.org - in particular, consider joining the SOCNET-L email list where a great group of people with deep SNA experience gather to discuss these kinds of questions.

My short answer is: no; I would not simply add those values together. The "key metrics (ie. Degree, Betweenness and Closeness centrality and Eigenvector centrality)" all mean very different things.

You may want to explore the ways these values change over time in the network. Each metric has a description that could help you interpret what changes in those values mean in your dataset (see: http://en.wikipedia.org/wiki/Centrality).

For example, increasing density for the whole graph (a metric that is reported on the Overall Metrics worksheet) suggests that more relationships are being created over the intervening time, while older relationships are being maintained. If a particular node's eigenvector centrality increases significantly, they may have created new important relationships or their friends may have developed new connections.

NodeXL can implement almost any model you decide to create by building formulas that relate the node metrics and other attributes. These compound metrics can be mapped to any display attribute using the Auto-fill columns feature.

Regards,

Marc

Nov 22, 2009 at 1:38 PM

Hello Marc,

Since my last posting, have located an interesting paper on Power in Networks (Valdis Krebs; http://www.orgnet.com/PowerInNetworks.pdf) which is very close to the topic that I wish to study and which appears to include an all-embracing metric for Power measurement. However, there is no explanation as to how the Power metric has been arrived at other than reference to the use of InFlow 3.1 software (http://www.orgnet.com/inflow3.html). Do you know what sort of formula or algorithm it uses and, if so, could it be replicated in NodeXL as this would be very useful?