Newbie: Help using nodexl for author relationship database

Jan 25, 2013 at 11:45 AM

I'm one-day using nodexl so I'm fairly clueless.

My need: I've a database that has: author last name, title of article, journal published, published date, office location. I'd like to use NodeXl (if feasible) to illustrate the relationships between authors, institutions, and location.

My problem: Not sure what column to make vertice(s) and which are edges?

I could add in columns if needed to give connectedness (?).

Any help is greatly appreciated!



Jan 25, 2013 at 2:30 PM

A network is represented in NodeXL as a collection of connections we call "Edges".

An edge is any pair of unique identifiers.

For your data, an edge is a pair of author names who have co-authored.

If there are three authors on a paper, there will be six edges - one for each pairwise combination of authors. (Unless order of authors matters, then there may be just three edges).

The edge may have many additional columns of associated data which could describe the connection (when did it happen, what journal, etc) or the entities involved (Age of author, Institutional Affiliation, etc.).

NodeXL will automatically import an edge list and place the data in the correct locations (edge data on the Edges worksheet, vertex data on the Vertices worksheet).



Jan 26, 2013 at 3:18 PM


Thank you for your help.

When I opened the database file, Nodexl asked me to identify the vertices and edges.

I'm thinking if I want to visualize relationship of author to location, author and location are vertices and other column data are edge related to the vertices?

Thank you,


PS. Also, when I do graph the data and then replot, the graph changes although I didn't change the data. Why does this occur?

Jan 26, 2013 at 5:14 PM


The core of network analysis is picking the correct entities and relationships.

If you have data about co-authorship and institutional affiliation (location) you could create many different kinds of networks.

Networks come in types depending on the different types of entities and relationships captured.

If all the entities are one type of thing, say people, then the network is a "unimodal" network.  If you have two kinds of entities (ex: people and papers) it becomes a bimodal network.  Three or more (ex: people, papers, places) and it is multimodal.

If all the connections among the entities are of the same type (ex: follows, co-authors, etc.), the network is a uniplex network.  If there are multiple ways the entities can connect (ex: follow, reply mention) then it is a multiplex network.

Both dimensions can be varied (ex: unimodal/multiplex, multimodal/uniplex).

Your job as network analyst is first and foremost to define the things that get into the population of vertices and the ways they can connect.

So, you could create:

Author <> Author (co-citation, unimodal, uniplex, network)

Author <> Paper (bimodal, uniplex)

Paper <> Paper (co-author, unimodal, uniplex, network)

Author <> Location/Institution (bimodal, uniplex)




Jan 26, 2013 at 5:18 PM

NodeXL provides two algorithmic layouts for placing each vertex in a location on the graph pane: Fruchterman-Reingold and Harel-Koren.

Both of these algorithms are non-deterministic: they create different outcomes every time they are executed.

If you want to lock the vertices once you have them in a location you like you can either select the padlock icon in the graph pane or select all vertices and change the Locked attribute to "Yes".  You can change selected vertex attributes by right clicking the graph pane and picking Edit selected vertex properties.