Page 1 of 1

Appendix 2: Social Network Analysis Measures

Evaluation of AHRQ's Pharmaceutical Outcomes Portfolio

Network SizeThe number of unique ordered pairs of actors within the network and is a basic demographic of a network. Network size matters, because it shapes the social structure of the network due to the capacity and resources needed to maintain relationships (Hanneman, 2000).a
Number of TiesA basic demographic of the network and is the count of the number of relationships or ties in the network (Hanneman, 2000). It can reveal how large and connected the network is which has implications for information and resource flows.
Average DistanceThe average number of relations in the shortest possible connection from one actor to another. Again, this metric has implications for information and resource flow; if distances are large it may take some time for resources to flow through the network (Hanneman, 2000).
DensityThe total number of actual ties divided by the maximum number of possible ties in the network (Kilduff and Tsai, 2003). It ranges from 0-100 and is the overall measure of connectedness of actors within the network. The higher the density of a network, the more connected the actors are which increases information and resource. Although the density measure of one CERT cannot be directly compared to that of another in part because the density is in part a function of the size of the network and would be meaningful comparison if networks were the same size. The density measure is difficult weight heavily because it is a somewhat artificial measure given the data collection did not include speaking with the partners in the network to ask who their partners given resource constraints, evaluation priorities, and the ego network framework. Therefore, naturally the density measure was lowered because of this for all the CERTs.
Degree CentralityMeasure of the ego actors' position within the network by counting the total number of direct connections of that actor. Core or central actors have many more connections than do those who are on the outside or periphery of the network. Those actors that are central within the network are in a position of power within the network (Kilduff and Tsai, 2003). The lower bound of this measure is 0 and its upper bound is a function of the total number of ties.b
ClosenessMeasure for networks that are fully connected and examines the "shortness" of the direct connections of the actor to other actors in the network. A large closeness measure positions the actor so that they can reach many other actors within the network, thus putting them in a power position within the network (Kilduff and Tsai, 2003). As with Degree Centrality, the lower bound of this measure is 0 and its upper bound is a function of the total number of ties.
BetweenessMeasure of an actor's ability to be a bridge or "go between" for other pairs of actors by being an intermediary connecting that relationship (Kilduff and Tsai, 2003). It ranges from 0-100; a high betweeness score signifies that an actor occupies a broker role within the network, and can mitigate contacts between other actors (Hanneman, 2000).
KeyPlayerMeasure resulting from a program that identifies the optimum sets of nodes to target for either removal or observation/intervention in a given network. Thus the key members of the network are identified and confirmed by core/periphery measures in Ucinet (Borgatti, Everett, and Freeman, 2002). A keyplayer plays a prominent bridging role in the network and its removal would result in a fragmented and less connected network. There are no upper or lower bounds to this measure, rather it locates key actors within a network based on data input. The norm is to locate 2-3 key actors, but that is a judgment based on the size of the network you are dealing with. The keyplayer algorithm is a metric designed to locate the main actors within the network diagram that if removed would fragment the network or in which their position in the network indicates an opportunity to expand the network.

a. Hanneman, R. (2000). Introduction to social network measures. Retrieved June 9, 2005, from Accessed December 20, 2007.
b. Kilduff, M., & Tsai, W. (2003). Social networks and organizations. Thousand Oaks: Sage Publications.

Return to Contents
Proceed to Appendix 3

Current as of December 2007
Internet Citation: Appendix 2: Social Network Analysis Measures: Evaluation of AHRQ's Pharmaceutical Outcomes Portfolio. December 2007. Agency for Healthcare Research and Quality, Rockville, MD.