Journal of College & University Student Housing Volume 48, No. 2 - 16

How Interactions Contribute to Thriving
cations, the institutions from the sample were all considered large (5000+ students)
research universities (seven designated as R1-very high research activity, and one as
R2-high research activity). Institutions were first selected on these two criteria (size
and research status) and then out of convenience (they employed administrators who
agreed to distribute the instruments). Seven institutions were public, and one was
private, and the mean enrollment from the institutions sampled was 36,239. Students
were asked to complete an online survey that consisted of items from the interactions
and thriving instruments, as well as demographic questions. Prospective participants
were offered an opportunity to enter into a random incentive drawing for completion
($25 Amazon card). Survey distribution was facilitated by institutional partners from
our personal and professional networks on these campuses who emailed LLC students
a survey link and invitation to voluntarily participate. After closing the survey, 1,322
total responses of all classifications (first-year through seniors) were documented.
After missing data and screening procedures, 903 responses were valid based on
completion. After removing all first-year students, the final usable sample for analysis
consisted of 206 sophomore, junior, and senior students (22.8% of usable surveys).
Information about the sample is offered in Table 1.
Data Analysis
This study employed a non-experimental design by examining the relationships
among variables (Sriram, 2017). The analysis was executed in two stages. First, measurement
models were created to appraise the integrity of all latent variables through
confirmatory factor analysis (CFA) techniques. Measuring the contribution of each
indicator toward its construct through CFA verified that the data measured the latent
construct it intended to measure. Performing CFAs through the maximum likelihood
method allowed us to test that all constructs met the level of statistical viability needed
for inclusion, as evidenced by acceptable fit indices.
After conducting the CFAs, we utilized structural equation modeling (SEM) as
our primary statistical technique. This approach assumed two central procedural
elements: the causal processes under study were represented by a series of structural
(regression) equations, and these structural relationships could be modeled pictorially
to enable a clearer conceptualization of the theory under study (Byrne, 2016). A
hypothesized model was created based on the review of literature that predicted how
interaction variables would be in relationship with thriving factors. After loading
data (N = 206) into SPSS 25 and checking normality and variance assumptions, data
was transferred into AMOS 25 and mapped onto the hypothesized model. Goodnessof-fit
measures at the following thresholds (Byrne, 2016; Kline, 2015) were utilized to
test model fit: CFI > .95, RMSEA < .06, and Chi-squared > .05 p-value.
RESULTS
Fit indices implied satisfactory fit for all CFAs. Factor loadings for all items also loaded
above the acceptable threshold of .40 (Matsunaga, 2010; Walker & Maddan, 2019).
Further, reliability measures (Cronbach's α) of all latent variables were suitable (> .80)
16 The Journal of College and University Student Housing

Journal of College & University Student Housing Volume 48, No. 2

Table of Contents for the Digital Edition of Journal of College & University Student Housing Volume 48, No. 2