***** To join INSNA, visit http://www.insna.org *****

Hi Noxta,


Glad to hear about your research.


What I would suggest is looking at A ties and B ties as two potentially overlapping (dichotomous) networks (this may or may not make sense depending on your participants). If communication occurs exclusively in A, it is a tie in the A network but not B. If it occurs exclusively in the B network, it is a tie in B but not A. If it is a mix, it is a tie in both.


Then, using XPNet, you can accomplish much of the analyses you're looking at in terms of reciprocity and triadic analyses. You can see if reciprocity is a significant tendency in one or both networks, or if various cross-cultural triadic configurations are more likely than others, as well as more hypotheses regarding activity and homophily.


There may be certain pairs of nodes for which a given language tie is impossible (because they don’t both know language B, for instance). For these, you would identify this as a ‘0’ in a structural zero file for the language which is not shared.


In terms of descriptive measures, I was you may wish to think of the network ties free from attributes first, and see if language-based attributes of the tie/node are predicted from things such as centrality.


Finally, may I ask what you mean by the term 'valued' here? Do you collect data on the frequency or intensity data, so that a tie might be strong or weak? Or by 'valued' are you referring to the fact that edges have attributes?






H. Colin Gallagher

Postdoctoral Research Fellow


Melbourne School of Psychological Sciences
Faculty of Medicine, Dentistry and Heath Sciences | University of Melbourne

12th Floor,  Redmond Barry Building, The University of Melbourne, VIC 3010, Australia


From: Social Networks Discussion Forum [mailto:[log in to unmask]] On Behalf Of Revotua Deu
Sent: Monday, 20 August 2012 8:45 PM
To: [log in to unmask]
Subject: Edge attributes / Multiple Networks / Signed Networks


***** To join INSNA, visit http://www.insna.org *****



Our data is about sociolinguistic relations of several teenage students (12 years old) in sociocentric networks of multiple schools. The sociolinguistic relations have two types of information: relations between adolescents and language used in this relation. Not all dyads have relations, and if relations exist, can be in language A or language B. Our information can be more complex, because language use is in language A, language B, language A and B, o other languages. But we can dichotomize the data.

We have problems with the analysis of our data, because edges have attributes, and and I couldn’t find any reference about how to handle whit it, neither with R nor any other software.

Then, I have three questions:

How can we analyse valued networks (qualitative links)? We are interested in multiple analysis at the descriptive level: language reciprocity, centrality in different languages, triad analysis with multiple languages relations, and so on...  We are working it combining UCINET results of full network, and network in language A and network in language B, and comparing it with Excel. Is it possible analyse better how reciprocity dyads have language reciprocity too? For exemple, “90% of reciprocity relations are in the same language”. Or 20% of transitive triads use A language in X -> Y and Y -> Z relations, and B language in X -> Z relation. We are also interested in more complex analysis like UCINET> Tools> Testing Hypotheses> Mixed Dyadic / Nodal> Categorical attributes> Relational Contingency Table analysis. And triad analysis with qualitative links (and node attributes).

How can we analyse edge attributes with ERGM? Is it possible? A possible way is work with multiple networks, like “relations in language A” and “relations in language B”. It is possible compare networks with ERG Models with XPnet. But our data is (theorically) negative correspondence, and we can’t interpret results of multiple networks analysis like it. Do you know bibliography that can help us?

Maybe a solution is working with signed networks. Origins of this analysis is other, about positive relations (friends) and negative relations (enemies). And balanced networks theory don’t adjust to language networks. But is not far that language use in networks can be understanded  like relations in A language (+), in B language (-) or in both languages (0). Wich software can work better with signed networks?

Last but not least, my presentation: I am Natxo Sorolla, and I am working in sociology of language at the University of Barcelona.

Natxo Sorolla


_____________________________________________________________________ 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.