Beginner Stuck on the Beginning

[scott.t...@gmail.com]

One, I am a complete beginner with ORA. Two, I have a bipartite spreadsheet (nodes x attributes), Three, I've read all the documentation and watched two videos. Four, however I am stuck the beginning. I can import my Excel file and designate a network name, designate Agent ( I want to analyze Agent vs Agent in social network approaches) but then I am asked to specify the attribute file and I specify the same file in all cases the first column is the ID column. Then ORA obliges at least the correct number.

At this point I was happy but not when analyses began as I saw no links displayed in overall visualization or in Sphere of Influence, i.e., just the nodes. I'm sure its some simple process step I am missing that is obvious but I typically search for more complex situations. I noticed that in the main window if I open my network i see only Agent. I cannot see how to designate the attributes as "knowledge."

I would appreciate clarity for this novice. Thanks very much.

[Mihovil Bartulović]

Hey,

Let me make sure I understand your problem correctly.

You have a spreadsheet that has two columns, and the data looks something like this:

Agent      Attribute
---------------------
Agent 1    Attribute 1
Agent 2    Attribute 2
Agent 3    Attribute 1
Agent 4    Attribute 3
Agent 5    Attribute 2
Agent 6    Attribute 1

Agent 7    Attribute 3

.

.

.

Is that correct?

If that is the case, you will have to do a bit of network manipulation to create an agent × agent network, given that your dataset has only agent × attribute information.

Let me know if I understood your problem correctly, and then we can go into more detail on how to create the network you need from the data you have.

Best,
Mihovil

[scott.t...@gmail.com]

Yes, you are correct. That is the input format i chose with Import Wizard. However, your comment suggests to me that ORA wants a symmetric(1-mode) matrix and a separate attributes file. If this is correct can I assume the attributes file will also have the same ID column? I assumed ORA wanted the full attribute file first as that is what the Wizard requested. Thank you.

[Mihovil Bartulović]

The issue is that ORA needs a network structure to create connections between nodes. If you provide ORA only with relations between an agent and an attribute, there is no way for ORA to connect the agents to each other.

Given your data, the only way to build a network of agents is to fold the network on itself.

Let’s say you import your data as a bipartite graph via:

Data Import Wizard → Import Excel or text-delimited files → Table of Network Links → give a name to your Meta Network.
Then, during the import process, you can select both your actor and attribute columns as Node IDs.

Let’s say we assign the actor column to the Actor class, and the attribute column to Knowledge (it could be anything else like Resource, Skill, etc.). It should look something like this:

In Step 3, we create the Agent × Attribute (Knowledge in this example) network by selecting source and target node IDs. After clicking Next → Finish, we’ll have a meta-network with an Agent × Knowledge bipartite graph, which might look like this in matrix form:

A =

        | Attr1  Attr2  Attr3
-----------------------------
Agent1 |   1      0      1
Agent2 |   1      1      0
Agent3 |   0      1      1

Now, if we transpose that matrix, we get Aᵗ:

        | Agent1  Agent2  Agent3
-------------------------------
Attr1  |    1       1       0
Attr2  |    0       1       1
Attr3  |    1       0       1

Next, by multiplying A × Aᵗ, we essentially multiply the Agent × Attribute matrix with the Attribute × Agent matrix, producing an Agent × Agent – shared attribute network:

         | A1  A2  A3
       ----------------
Agent1 |  2   1   1
Agent2 |  1   2   1
Agent3 |  1   1   2

This network now represents Agent × Agent relationships, where the link weight corresponds to the number of attributes two agents share.
For example, if the attribute is “Knowledge” and represents languages spoken, and Agents 1 and 2 both know English and Spanish but not French, their link weight would be 2. If two agents know all three languages, their link weight would be 3.

This is the only way to create an Actor × Actor network from the kind of data you have.

________________________________________________________

Folding the graph in ORA:

After importing the network, you should see something like this:

• Right-click on the Agent × Knowledge network → Fold
• A menu will pop up asking whether you want a Row Fold or Column Fold
• Select Row Fold → Click Fold

ORA will immediately create a new folded network under the same meta-network. You can use this folded network for your analysis.

However, be mindful that the link weight represents shared attribute count, and it’s not as straightforward as a typical 1:1 relationship between two agents.

__________________________________________

If your data looks like this instead:

Agent Source    Agent Target  
-----------------------------------------
Agent 1         Agent 2      
Agent 2         Agent 3     
Agent 3         Agent 4        
Agent 4         Agent 5     
Agent 5         Agent 6      
Agent 6         Agent 7     
Agent 7         Agent 1     
...             ...             ...

This format already contains direct links between actors, and ORA will immediately recognize it as a one-mode Agent × Agent network. You can proceed with your analysis without needing to fold anything.

But given the data you described earlier (Agent × Attribute), folding is the only way to derive an Agent × Agent network.

Hope this helps!

[scott.t...@gmail.com]

Ok. I got most of the way along the example but I have 205 attributes. With the example, I for each column I selected NodeID and Knowledge. It took a fair amount of time (too bad noteclass could not be specified back at the spreadsheet level say as an extra row or special heading). Anyway I cannot select the Target as an Attribute even though I copied your example. All I get are 205 original column labels so I am still missing something. However, I really like the ability to have several nodeclasses in a dataset as the results could prove illuminating. Thanks,.

[Mihovil Bartulović]

Do you have 205 unique attributes in a single attribute column, or do you have 205 unique attribute columns, each with different values?

If it’s the former, then you can proceed as I described earlier. However, if it’s the latter, you can create an Agent × Attribute₁ network all the way to an Agent × Attribute₂₀₅ network, and you can fold each one individually to examine relationships between agents based on shared attributes.

For example, in a company setting, folding an Agent × Knowledge network to create an Agent × Agent shared knowledge graph is very useful. If someone is out on vacation, you could easily identify if someone else has similar knowledge and can cover for them if needed.

Hope that makes sense.

Best,
Mihovil

[scott.t...@gmail.com]

It's the latter. So I'll try that next. Thanks.

[scott.t...@gmail.com]

Still not working. The first window (Import Wizard) asks for the format of the Excel file and I chose an agent x attributes like the example shows (Image 1), then an image nodeset Knowledge (Image 2), but then in Image 3 it ORA requires input of the attribute nodeset. Still confused as I would think Agent would be input first and then input the class chosen. Thank you,.

[Mihovil Bartulović]

Your steps are for importing attributes so they can be appended to already existing nodes—i.e., you already have an agent × agent network from a different dataset, and now you want to add, for example, demographic data to each of the agents so you can filter using those attributes.

The method I was describing uses Data Import Wizard → Import Excel or text-delimited files → Table of Network Link

[scott.t...@gmail.com]

Success. No longer stuck now that I see the light. Basically, ORA receives one or more matrices where the columns and rows are specified as a specific category, and one can keep adding matrices based on their category. Now, I have to learn the specifics, e.g., node labeling in visualization, etc.. Thank you very much.