4 ms·
Looks impressive, but a simple N x K contingency table would help understand the data better. Here, N=K=15 and entries of the table would be raw counts (e.g. nu
by ramanujan 14y ago
Looks impressive, but a simple N x K contingency table would help understand the data better. Here, N=K=15 and entries of the table would be raw counts (e.g. number of major i in career j).
You could color the cells by a few different criteria, e.g. absolute scaling, row normalization, and column normalization with red for "lots" and blue for "few". Maybe toggle those three criteria via a radio button.
The advantage of such a visualization is that it allows you to see both trends (which rows/cols are more red, and which blue) and specific numbers. It is also immediately interpretable for a new viewer without any explanatory preamble. Sometimes the simple stuff is best.
- bicknergseng 14y agoI would say the advantage of the original visualization is the clear lines drawn between n and k. The table is very complex without filtering, but imo very clear when you mouse over the majors on the right. I would say the real problem is that the subsets are very uneven; the orange group dwarfs the other two, which arguably makes it more difficult to understand the trends. Along that line of thought, however, I would argue that the Psych majors belong with the green group.
- aetherson 14y agoAs a Williams alumnus: the major color groups correspond to the "divisions" of courses at Williams -- the administration groups all departments into three divisions, and requires that students take a certain number of classes from each division. So the chart faithfully reproduces the decision of the Williams College administration to include Psychology in Division II (Social Sciences) rather than Division III (math and hard sciences).