I would like to survey a large ego-centred network in my questionnaire (n = 25). However, this large number of alters results in a very high number of questions concerning the alter-alter relations (n = 25 results in 300 possible relationships). So querying these relations individually, as would be possible in the Dyad Census Interface, is very time-consuming for the respondents.
To reduce the respondent burden, I would therefore like to use the alter-linking approach. In this approach, the alters are displayed one after the other in the middle of the screen (in a Sociogram) and the respondent is asked to indicate his or her connections to the other alters displayed. So I imagine the Sociogram interface with automatic positioning of the nodes, except that one node is in the centre and there is a sociogram for each node.
Is there a possibility to implement this approach in Network Canvas?
Thanks for the question! We were able to chat about this recently at our weekly meeting.
Indeed, part of the reason for annotating links on a sociogram is because going through a dyad census of that length can be pretty onerous. On the other hand, we appreciate that doing it in a sociogram in its current form may lead to false negatives,. especially in densely connected pockets of nodes.
At present, the dyad census and the sociogram are really the only two ways forward for collecting edge data. A one node to many potential nodes sort of interface is a great idea on the horizon but we cannot say whether it will be included in future iterations of Network Canvas.
In the meantime, here’s a slightly complicated way to help that might work in some circumstances:
Denote which group nodes belong in (like all friends / all family or all hometown, etc…) and then only do a dyad census within each group using node filtering.
Present the results on a sociogram with automatic layout. Since there should be connectivity within the groups, they will all cluster in different corners of the layout.
Then add group-spanning links on the sociogram.
Why might this work? Most ties are clustered in groups. We use a dyad census to cover all possible relations, but in practice there are obvious clusters where we are likely to see relations and many places where we are not likely to see relations and observing one might be a fluke.
Why this might not work? It might be hard to partition the nodes into separate groups without ending up sometimes back with a primary group with many many ties. Also, it still might be hard to see edges between groups or recall specific links on a sociogram.
I hope that helps. Thanks again for the idea and the comment. Best of luck on your research!
one thing that you might consider is labelling nodes in groups and then only doing a dyad census within group. That is a census between all family or one for all friends. It’s a bit tedious but it would filter down nodes. Group-spanning edges would be absent, but if you do an automatic layout, each group will have it’s own ‘oribit’ on a sociogram, and so perhaps you can then do group-spanning edges on that sociogram.
