journal_vol49no2 - 60

residence halls had the lowest room change rate at 1.6%. This second group of students
(n = 3,182) is represented as a terminal node, not further differentiated by room
change request percentage in Figure 1. The final group is those first-year students that
lived in the double-corridor luxury residence halls (n = 6,728) who changed rooms at a
rate (2.9%) mostly consistent with the overall room change rate (2.9% vs. 3.1%). These
results are generally consistent with previous design and student interaction research,
except for the DC: Centralized residence halls (Brown et al., 2019). These results suggest
that another factor beyond socializing design may impact room change behavior
for first-year students in these residence halls.
The second room change differentiator in the decision tree was first-semester GPA
for both the Suite/DC: Centralized and Hybrid Luxury branches. High academic performers
in both branches were less likely to change rooms. For the Suite/DC: Centralized
branch, those students with a GPA of above 3.65 changed rooms less frequently
than did other students (2.4% vs. 4.5%). The study variables did not further differentiate
the non-high performing students (n = 5, 603) in their room change rate. For the
Hybrid Luxury branch, those students with a GPA of above 3.47 changed rooms less
frequently than did others (1.9% vs. 3.6%). Both subsamples were further differentiated
based on the study variables: race difference for the non-high performing sample
and age for the high-performing sample.
Of the four GPA subsamples defined in the second level of the decision tree, three
were further differentiated on room change requests. First, the room change requests
for high performers in the Suite/DC: Centralized branch were differentiated based on
roommate differences in AGI. The number of room change requests for those having
medium (6.8%) and very large (4.2%) differences in AGI was greater than expected
based on the parent node change rate (2.4%). Those with missing values and small
differences in AGI were less likely to change rooms (2.0%) than what would be expected
for high performers. Of the 276 high performers with large AGI differences,
only one first-year student changed rooms. Although these cut-points were statistically
significant, the smaller cell sizes may indicate model overfitting. Second, for the high
performers (n = 509) in the Hybrid Luxury branch, 18-year-old students were much
less likely to change rooms than were 19- and 20-year-olds (1.6% vs. 3.5%). Given the
smaller proportion (16.5%) of the older students, this may be driven by differences in
other factors related to age, such as maturity, rather than age itself. Finally, the nonhigh
performing subsample in the Hybrid Luxury residence halls was further divided
by differences in race. Residents with roommates of a different race were much more
likely to change rooms mid-academic year (4.5%) than were students rooming with
someone of their same race (2.9%).
Residence hall design was the strongest predictor
of room change requests.
Volume 49, No. 2 * 2023 57

journal_vol49no2

Table of Contents for the Digital Edition of journal_vol49no2