***** To join INSNA, visit http://www.insna.org *****
First, a reminder that there is a UCINET user's group / listserv that
questions like this can be directed to. Since not everyone on this list uses
UCINET (!), it seems to me polite not to direct questions about UCINET to
Second, degree centrality and degree centralization are conventionally
defined for non-valued data. There is no one, authoritative, way to
calculate these concepts for valued data. Ucinet chooses to interpret valued
degree as the sum of tie strengths for each person. In current versions of
UCINET, degree centralization for valued data is calculated by defining the
most centralized valued graph as a star in which the central node has
maximum-valued ties with all others, and all others have ties only with the
center. This yields a valued centralization score that varies between 0 and
100 percent, which eases some people's minds.
Older versions of UCINET were not willing to affirm this particular
normalization, and simply calculated centralization as if the center had
ties of strength 1 to all others. This could yield values greater than 100%.
To convert between these older values and the new centralization values,
simply divide the centralization that old ucinet reported by the maximum tie
strength in the network.
Please note that other approaches to valued degree centralization are easily
imaginable and UCINET's recent adoption of one approach should not be
regarded as definitive.
Third, the distance-based cohesion measure is basically the sum of
reciprocal distances, with the convention that the reciprocal of infinite or
undefined distance is 0. It is normalized by dividing by the number of
directed pairs. It is a way of looking at average distance that works
reasonably in unconnected graphs. I use it in my keyplayer work (see
http://www.analytictech.com/borgatti/papers/cmotkeyplayer.pdf ), based on a
recommendation by mark newman. Bonacich and Bienenstock have independently
used it in similar work. But it has been invented many times in the
Org Studies, Boston College
[log in to unmask]
>From: Social Networks Discussion Forum [mailto:[log in to unmask]] On
>Behalf Of Siyuan Huang
>Sent: Tuesday, March 01, 2005 11:59
>To: [log in to unmask]
>Subject: [SOCNET] questions about network structural properties
>***** To join INSNA, visit http://www.insna.org *****
>I am a graduate student working on a social network project. I ran into
>questions about social network structural properties when I was using
>to analyze network data.
>1. network--->centrality--->Degree: The output provide Freeman's graph
>centralization measures. I understand that the degree centralization
>expresses the degree of inequality or variance in the network as a
>percentage of that of a perfect star network of the same size. The output
>also indicates that centralization of a network could exceed 100% if valued
>data are used. My question is how should a network centralization value
>that exceeds 100% be interpreted? What does it mean? Does it mean the
>network structure is more than a perfect star structure?
>2. network--->cohesion---->Distance: the output provides distance-based
>cohesion. It is also indicated that this value ranges from 0 to 1, with
>larger value meaning higher degree of cohesiveness. How is this value
>calculated? What exactly does it mean?
>I appreciate any feedback from you! Thanks!
>SOCNET is a service of INSNA, the professional association for social
>network researchers (http://www.insna.org). To unsubscribe, send
>an email message to [log in to unmask] containing the line
>UNSUBSCRIBE SOCNET in the body of the message.
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.