Making a graph look neater overall

Aug 15, 2013 at 6:41 PM
I recently downloaded NodeXL and love the program but on the dataset I am currently working on it seems like there is no good way to preserve all the connections between the vertices and at the same time make it easy to follow. I think the problem is that the vertices are too inter-connected with each other and that there are no real outliers creating lots of cross traffic. I have tried different layouts but none of them really helped and groups worked a little but it really just moved the vertices around instead of clearing up more space. I should mention that the data I'm using has around 30 vertices and they are all categorical data. Does anyone have any advice on how to make a graph like this more readable? My graph looks very similar to the ones on this webpage but none of the tricks the author used can really be applied to my grpahs it seems. Very Similar Graphs
Aug 15, 2013 at 9:23 PM
Can you share the workbook?

I can see if I can get a better result.

-- Marc

Sent from my iPhone

Aug 15, 2013 at 9:35 PM
Sure what would be the best way to get that to you? I would try to put it here but I do not see anyplace to attach files.
Aug 15, 2013 at 9:45 PM
You can upload to the NodeXL Graph Gallery via the export menu.

Or attach it to an item in the issues list.

-- Marc

Sent from my iPhone

Aug 15, 2013 at 9:50 PM
Edited Aug 15, 2013 at 9:52 PM
Ok I just uploaded it to the gallery. Its called Resource Management Strategies.
Aug 15, 2013 at 9:58 PM
Try the circle layout and use a metric for layout order that is useful: maybe degree or another attribute value (head count, budget, etc)?


Sent from my iPhone

Aug 15, 2013 at 10:29 PM
I have tried playing around with that now and haven't seen much better results unfortunately. I have exported my best looking attempt to the gallery and labeled it Resource Management Linkage.
Aug 16, 2013 at 2:01 AM
Edited Aug 16, 2013 at 2:07 AM
Aug 16, 2013 at 2:04 AM
With more data about the entities and their connections, the graph can be made to reveal more about the structure of the network.

Do you have weights for the strength of the connection - the amount of something they exchange?

Do you have attributes that would group these entities - which divisions are these entities in together?
Aug 16, 2013 at 5:05 PM
First off the grpah is looking much better already thank you for doing that.

For the strength of connection the data is pretty generic its either strongly connected, weakly connected, or not connected at all. I tried to represent this by having strong connections as solid lines and weak connections as dashed lines. Other that that there is no real inormation about how they connect.

As for grouping the verteicies follow this outline. (group # name: verticies in group)

Group 1 Reduce Water Demand: Agricultural water use efficiency, Urban water use efficiency

Group 2 Improve Flood Management: Flood management

Group 3 Improve Operational Efficieny and Transfers: Conveyance-Delta, Conveyance-Regional/Local, System Reoperation, Water Tansfers

Group 4 Increase Water Supply: Conjunctive Management & Groundwater, Desalination-Brackish & Seawater, Precipitation Enhancement, Recycled Municipal Water, Surface Storage-CALFED, Surface Storage-Regional/Local

Group 5 Improve Water Quality: Drinking Water Treatment and Distribution, Grounwater/Aquifer Remediation, Matching Quality to Use, Pollution Prevention, Salt & Salinty Management, Urban Stormwater Runoff Management

Group 6 Practice Resource Stewardship: Agricultural Land Stewardship, Ecosystem Restoration, Forest Management, Land Use Planning & Management, Recharge Areas Protection, Sediment Management, Watershed Management

Group 7 People and Water: Economic Incentives, Outreach and Education, Water-Dependent Cultural Resources, Water-Dependent Recreation
Aug 16, 2013 at 5:36 PM
You can group nodes (we call them "vertices") together in a number of ways.

Directly from the graph pane you can drag vertices that should be in the same group together. The selection tool (or Control-Click) allows you to select more than one vertex at a time.


Then right-click the graph pane (any white space will do). Select the "Groups..." menu option:


Select the option, in Groups, for Add Selected Vertices to Group:


Provide a name for the Group:


Repeat this process for each group you would like to create.

An alternative method involves adding this group membership attribute to the Vertices Worksheet as a new column (see the column "Add Your Own Columns Here"). Then, it is a simple matter to use the NodeXL >Analysis>Groups feature to assign each vertex to the proper group.

If you do get your graph grouped, you can then apply the "Group-in-a-box" feature from the Layout Options menu in the Graph Pane layout menu. This feature will locate each set of vertices that share a group in a separate box (like a "Bento box" for networks).

-- Marc
Aug 16, 2013 at 6:36 PM
I posted the results of what you suggested (Resource Managment Grouped), but I cannot figure out how to keep the labels big while reducing the sizes of the edges as you did previously. But even with that it looks like there is still a lot of a cross traffic between the verticies. I thought about combining those edges in the layout options, menu, but when I tried it I noticed that I will lose some information about how the verticies are connected to other vertices. I realize at a certain point there may be nothing that can be done about the cross traffic but its already looking much neater than before.
Aug 16, 2013 at 6:47 PM
Edited Aug 16, 2013 at 6:48 PM
You can adjust the scale slider to change the vertex size.
You can also select Graph Options from the graph pane and select Other and Font to set the label sizes.
You can also turn off the edges that are not the strongest edges.
In the end, however, a nearly completely connected graph is a nearly completely connected graph!

-- Marc
Aug 16, 2013 at 9:32 PM
Thanks so much those really helped clear it up. Do you have any further suggesstions that I could try. I think as it is it looks pretty good but I am just wondering what I can do to make it look better.