2 - modes networks and trcikle down options

Feedback & Feature Requests
1

Hi,

I am trying to create a flow in which ego identify alters, then ego identify places, then ego matches alters to locations (many-to-many). Any ideas? (Important, none of the lists is fixed at the start of the interview).

2

Hi! Welcome to the community :slightly_smiling_face:

I think this is possible to approximate, although there is one important limitation.

Unfortunately, there isn’t currently an interface that will enforce exactly the workflow you describe at the matching stage. The Dyad Census and Tie-Strength Census operate on a single node type and consider all pairs of nodes shown on the stage. Network filtering can control which nodes appear, but it cannot express a rule such as “show both people and places, but only ask about or allow ties between the two groups.” This is similar to the limitation discussed in an earlier thread about two-mode networks.

The closest workaround I can think of would be the following:

  1. Create a single node type, perhaps called Entity, rather than separate Person and Place node types. This is necessary because the Sociogram displays nodes of a single type.

  2. Use two Quick Add Name Generator stages. The first would ask ego to name people and automatically assign an additional variable such as is_person = true. The second would ask ego to name places and assign is_place = true. Quick Add supports assigning additional variables to nodes as they are nominated, so both lists can be created during the interview—you do not need a fixed roster or starting list.

  3. Show all of these entities on a Sociogram. I would probably use a custom background with “People” on the left and “Places” on the right, and first ask the participant to arrange the nodes into the two groups. The Sociogram supports custom backgrounds, participant positioning of nodes, and the creation of one or more edge types.

  4. Use a second Sociogram prompt to create the matches. For example:

    Connect each person to every place with which they are associated. Please do not connect two people or two places.

This will allow a many-to-many structure: a person can be connected to several places, and a place can be connected to several people.

The main drawback is that the software will not prevent the participant from accidentally creating a person–person or place–place edge. You would need to make the two groups visually clear, give explicit instructions, and check the exported edge data for invalid connections.

Finally, if you want to ask follow-up questions about each person–place match—for example, how often the person visits the place—you could add a Per Alter Edge Form after the Sociogram. That interface will present a form for each edge that was previously created, which can contain any attributes you like.

One other thing to consider is whether you need ego to explicitly review every possible person–place combination. The Sociogram is a good way to record positive matches, but it does not confirm that every possible pair was considered. If that is an important requirement for your study, then I don’t think the existing interfaces can provide a true two-mode census without some custom development.

Lastly, we have a new feature launching soon which will make this a little bit easier: custom node shapes, including mapping to attributes. This would allow you to define the single node type discussed above, but then map an attribute to the shape so that, for example, people render as circles but places render as squares. If you’re interested in trying this out I suggest keeping an eye open for a big release announcement in the next couple of weeks.

Hope this helps!

1 Like