Two levels of representation?

Nov 20, 2013 at 6:11 PM
Hi, I can't find a scenario that matches my data in the NodeXL manual. My data consists in small group discussions between researchers and teachers. A preliminary look at the data revealed that there was no direct contact between the two groups (participants seemed to mostly contribute messages that were not addressed to someone in particular). If I just do a directed vertex-to-vertex mapping, the graphs will be quite useless, showing very limited connection. My reading of the data showed that the conversation was centred around topics and not necessarily around a 1-on-1 exchange between participants. For example, a week-long discussion can be centred around 3 topics taken up by different participants. My question is how can I do a two level representation that shows (1) all the users that participated each week, (2) the in-between participants interactions and (3) the focus of each posting. In other words, I want to show that user A posted a message that is not addressed to someone in particular and in that message he addressed topics 1 and 2...and also show that user B did the same, or replied to user C and dealt with topics 1 and 2.
Thanks for your help
Nov 20, 2013 at 6:22 PM
It sounds like you want to construct a two mode network of people and topics rather than a one mode network of people linked to people.

In this case, replace Vertex2 with the name of a topic instead of the name of another person.


Nov 20, 2013 at 6:26 PM
Thank you for the prompt reply Marc. That's what I wanted to do first but that does not allow me to map the person-to-person interaction. If I put user A in vertex 1 and Topic 1, 2, 3 in Vertex 2--where can I include the information that user A also replied to point raised by user B while discussing topic you see my point?
Nov 20, 2013 at 6:35 PM
You are describing two different networks.

You *can* mix these edges, creating a multiplex as well as bimodal network.

Add an additional column to the edges worksheet, call it "Relationship Type".

You could create edges like:

Person1, Person 2, Replies-to, Date, Topic list
Person1, Topic 1, Mentions, Date, Person List


"Topic List" is an list of terms that describe a reply from Person 1 to Person 2.
"Person List" is a list of people names that have received a message from Person 1 containing the topic term.

You can visualize this network, but metrics generated on it will be of ambiguous meaning as you are mixing different types of entities (people and topics) and different kinds of edges (replies-to and mentions).

Warning: Things will get complicated if a Person can also be a Topic!

Hope that helps!