I recently developed some code to take student enrollment information and convert that into a visual map of the curriculum, showing how enrollments flow from one course to another. For example, you'd expect a lot of BIO 101 students to take BIO 102 within the next two semesters. In order to 'x-ray' course offerings, I have to set thresholds for displaying links. For example, a minimum transfer of 30% of the enrollment from one course to another in order to show up. There are many ways to add meta-data in the form of text and color, for example using the thickness of the graph edges (the connecting lines) to signify the magnitude of the flow. This is a directed graph, so it has arrows you can't see at the resolution I've provided. Other data includes course name and enrollment statistics, and the college represented. It can be used to isolate part of the curriculum at a time to get more fine-grained graphs.
In the graph below, it's a whole institution's curriculum. The sciences, which are highly structure, clump together in the middle. Less strongly linked structures are visible as constellations around the center. I particularly like the dog shape lower left. This sort of thing can be used to see where the log-jams are, and to compare what advisors think is happening to what actually is.
That's pretty intriguing. What did you use to create the map?
ReplyDeleteIt's a Perl script that takes registration information as input, does a lot of sorting and averaging, and uses the open source package GraphViz (with the perl package as glue) to create the graphs. I also use GraphViz to map correlations in big data sets. See this article, for example: http://highered.blogspot.com/2011/12/x-raying-survey-data.html
DeleteInteresting, I've been doing a similar project with the newly released ACGME Milestones for graduate medical education. My goal is to empirically justify a rationale for centralized assessment while at the same time discerning which milestones should be centralized. Once all of the milestones are released I anticipate having to link over 1800 different competencies across 26 medical specializations. I've been using Gephi and adjacency tables on just 4 specializations, which has yielded 397 nodes and 145 edges.
ReplyDeleteI remember you had mentioned this earlier, Dave, but I had not seen it. Right now this is bassd on conventional enrollemt data. I would be curious to see what could be visualized with the more complex data that academic institutions can gather using eLumen, tracking individual students relative to expected student learning outcomes/capabilities/competencies.
ReplyDeleteDavid, this is something we could collaborate on if you want. If you've got the data, I can modify the program.
DeleteMore! More! As a user of eLumen, I can say that this and other similar ways to visualize emerging evidence woul be most useful!
ReplyDelete