• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

A.Semenov held a public seminar on network analysis

On April, 8, 2011 a public seminar on network analysis was held in the Laboratory. It was delivered by Alexander Semenov, Ph.D. student at HSE in Moscow. The seminar was organized to study the basics of ERGM (p*) – Exponential Random Graph Models.

On April, 8, 2011 a public seminar on network analysis was held in the Laboratory. It was delivered by Alexander Semenov, Ph.D. student  at HSE in Moscow. The seminar was organized to study the basics of  ERGM (p*) – Exponential Random Graph Models. These models are one of various methods of analyzing social networks.

Why should we use statistical approach to analyze social networks? Alexander Semenov says that there are several reasons.

First of all, this method allows solving descriptive tasks; in particular it describes numeric characteristics on different levels such as nods, dyads, triads and networks in general.

The second reason is that generative tasks are solved. We can explain how the network was created, what the local basis of global network structure is and what the hidden processes of network shaping are. Having a graph only we can assume how it appeared. We also study whether there are trends of linking, clustering or centralizing; and which social processes influence on network.

ERGM is convenient as we can include several variables into the model simultaneously. These can be not only such parameters as age, gender and race, but structural effects as well. These can be reciprocity, transitivity, etc. Using this technique we can study which effects have the biggest explanatory significance. One more advantage of ERGM is that it is compatible with several other programs, such as Pnet, Statnet and StocNET.

One of the tasks was to study a scheme from Pnet manual that showed parameters and structural effects of the networks. The audience has also learned that structural characteristics of a network often influence more on network types that biological species. As an example Alexander used a map of social networks between chimpanzees and macaques. 

Practical implication of ERGM was demonstrated on the research of Alexander Semenov, who studies networks in livejournal.com community. He took 4 types of networks and analyzed the effects in and between them by ERGM. During the practical part of the seminar the participants learned the basics of ORA and its numerous options for visualization of network data. One can anonymize a network, download pictures of agents etc. ORA can be also used for calculating R, R2, and other statistic parameters.

Alexander Semenov swept his audience along him and the participants were really impressed by his seminar.

By Ksenia Medvedeva