*****  To join INSNA, visit  *****

Dear All,

I'm carrying out a social network study on language use in an electronic community. Data are drawn from the email archive containing all the messages exchanged among community members.

The study hypothesizes a positive relationship between network similarity and  similarity in language use.

I organized my data as following:

-First, I divided the archive into 60 topic-specific email subsets (groups of emails on the same topic);

-Second, for each of the 60 email subsets I built a two-mode matrix (ROWS = community members which sent or recieved at least one email on that topic; COLUMNS = email sent on that topic).

-Third, I computed with UCINET VI a meaure of similarity among the columns of those 60 two-mode matrices. So, I got 60 square mail-x-mail matrices, where xij = value of network similarity between email i and email j. I call those matrices "Network-Similarity Matrices".

-Fourth, I have other 60 square mail-x-mail matrices (one for each topic-specific subset), where xij = value of similarity between the TEXT of email i and the TEXT of email j. I call those matrices "Text-Similarity Matrices".

Now, in order to test the relationship hypothesized above, I would like to do the following:

-Building two diagonal matrices. The first one should have all the "network similarity matrices" on the diagonal and structural zeros elsewhere. The second one should be exactly the same with the "text similarity matrices" on the diagonal.

-Run a QAP regression using those two big diagonal matrices as inputs.

May I kindly ask you an opinion on these last two steps of my analysis? Do they make sense to you and what kind of weaknesses do you notice? Do you know other studies adopting a similar approach?

Thanks a lot,

Best Regards,

JosŤ De Fatima Garrois


SOCNET is a service of INSNA, the professional association for social
network researchers ( To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.