Entry 14 The Migration of College Students among their Declared and Chosen Majors – Visualization and Causal Analysis
Jun Wang, Klaus Mueller — Stony Brook University

The two chord diagrams demonstrate the migration of ACT-tested students among majors at different times. Chord width and color represent the number of migrating students and origin, respectively. Migrations <5% of a major’s population were filtered out for clarity.

The network shows the significant causal relations between all pairs of variables. We replaced the original variables “T{1,2,3}_Level2” and “HighestLevel2” more descriptively with “Migrated {Plan-Y1, Y1-Y2}” and “In Best Fit Major {Plan,Y1,Y2}”. Details on the causal structure discovery are presented in our upcoming VAST 2015 paper “The Visual Causality Analyst”.

The chord diagrams reveal that there was significant migration before Y1 but not before Y2, with some majors affected more than others. For example, almost no student declaring an interest in Health Administration & Assisting in high school actually went to it. Similar is true for Health Science & Technology which loses many students to the basic Sciences. Conversely, Business is the hottest major – many choose to switch to it and this trend continues in the second migration.

The casual network captures all significant statistical findings in one graph. Some of these are:

  1. The bi-directed edge between nodes N6 and N7 means that besides high ACT-scores there are also other possible causes for students to go to a 4-year college.
  2. A high ACT-score makes transfers less likely and 4-year college more likely. (3) Gender influences the ACT score (N1) – males score higher.
  3. Higher-scoring students exhibit better a judgement of their capabilities – students with high ACT scores (N2) have a better fit for their planned majors (N3, N11).
  4. Now those students (N11) have two choices – migrate to another major (N9) at lower likelihood or stay in the same major (N12) with higher likelihood. The network shows that the former group never really find the best-fit major again (N4, N10, N13, N5), while the latter group stays in the best fit major for the remaining years (also path N11, N12, N13).
Jun Wang, Klaus Mueller
E-mail: {junwang2, mueller}@cs.stonybrook.edu.
Visual Analytics and Imaging Lab, Computer Science Department, Stony Brook University
Stony Brook, NY 11790.