2.6 Studies Two and Three: Quantitative, Cross-Sectional Research
2.6.5 Data Analyses
2.6.5.1 Systematic Review.
Data for Study One were studies collected following a comprehensive and rigorous search of data bases. The quality of studies, which met the eligibility criteria for inclusion, was critically appraised. Appraisal consisted of an examination of each study for risks of bias (ethics approval, funding, publication, and geographical location), and for methodology (study design, demographics, types of student residence, and measures of social support).
A further appraisal was conducted of homogeneity and heterogeneity across the studies for validity of results.
Research Questions One and Two were investigated using findings and conclusions of the selected studies.
45 2.6.5.2 Study Two.
Data collected for Study Two were compiled and analysed using IBM SPSS Statistics
(Version 26) predictive analytics software. The alpha level to determine significance was .05 for all statistical tests.
Descriptive statistics were found, namely, means, medians, and standard deviations for continuous variables, and frequencies and percentages for categorical variables. Cronbach’s alpha measured the internal consistency of scales and sub-scales to assess reliability. Research Question Three was investigated in two ways, first, using one-way analyses of variance (ANOVAs) to find any differences between types of residence in levels of the continuous predictors, perceived available social support and domestic support. Second, cell sizes were obtained for the categorical variables, English use and Greeting network. Cell sizes did not meet the assumption (n ≥ 5) for Chi-square tests of difference. Analyses proceeded to non- parametric Spearman’s rank correlation coefficients to determine the strength and direction of the linear relationships between all continuous and categorical predictors and outcomes.
Research Questions Four to Seven employed linear multiple regression analysis to find effects of the continuous factors (perceived available social support, domestic support, meaning-in- life presence, and meaning-in-life search) on the variance in pass rate, intention to persist and psychological distress. Logistic regression analysis examined effects of the categorical predictors (greeting network and English use) on the variances of outcomes.
2.6.5.3 Transition from Study Two to Study Three.
To determine the direction of Study Three in 2020, Study Two data were further analysed. The analysis sought indicators which could explain higher perceived available social support in colleges than in flats/houses or “villages”. Means of the ISEL sub-scales were calculated. Individual sub-scales were entered into a correlational matrix (tangible, appraisal, belonging, greeting network, meals provided, total domestic support, meaning presence, meaning search, and mental health). Cohen’s (1988, pp. 79 - 81) guidelines
46 (small rs = .10 - .29, medium rs = .30 - .49, large rs = .50 - 1.0) determined the strength of correlations. Belonging had a stronger correlation with mental health than did either perceived available tangible or appraised social support. Moreover, belonging was moderately
correlated with greeting network, meals provided, domestic support, and meaning presence, as well as psychological distress, forming a cluster.
Literature related to belonging encompasses research on sense of community and the work of Manning (1994) in the field of military psychiatry. Relocated first-year
undergraduates were compared with typical military recruits and their shared characteristics identified. Both groups are within the emerging adult stage of lifespan development (aged 18 - 25 years) (Arnett, 2004); leave family homes, familiar communities, and relocate; adapt to a different lifestyle from that of home; become members of a small entity (student
residence/military unit) and large entity (university/defence force), face greater psychological challenges than their peers (urban/civilian); and rely on developing bonds of belonging and sense of community with peers to sustain them psychologically.
Manning (1994) found four key characteristics which engender and maintain socially supportive bonds (“cohesion”) between members of small entities and the increasingly larger entities in which they inhere. Such characteristics are material comfort (especially shared meals), ease of communication with those in authority (vertical) and peers (lateral),
commonality of purpose, and interpersonal care. In short, the construct consists of comfort, communication, commonality, and care. The author uses the term “integrated support” for the construct and the novel Integrated Support Scale (ISS) for the measure.
2.6.5.4 Study Three.
Data collected for Study Three were compiled and analysed using IBM SPSS
Statistics (Version 27) predictive analytics software. The alpha level to determine significance was .05 for all statistical tests.
Descriptive statistics were found for all variables, means and standard deviations
47 for continuous variables, frequencies and percentages for categorical variables. Cronbach’s alpha measured the internal consistency of whole scales and sub-scales to assess reliability.
Research Question Eight was investigated in two ways, first, using one-way analyses of variance (ANOVAs) to find any differences between types of residence in levels of the continuous predictors, integrated support, belonging, and sense of community. Differences prompted further investigation.
A matrix of Spearman’s rho correlations was compiled to determine the strength and direction of linear relationships between all continuous and dichotomised categorical
variables. Linear regressions of integrated support on belonging and residence-sense of community followed.
Research Questions Nine to Fifteen included continuous variables (belonging, sense of community, estimated GPA, and well-being) and categorical (intention to persist, regular contact with dispersed, distal co-residents and university peers). Significant differences were sought between types of residence using ANOVAs for continuous variables and Chi-square tests for categorical variables. Cell sizes did not meet the assumption of (n ≥ 5) for Chi-square tests of difference. Regression analysis followed. Multiple linear regressions sought effects on variances of continuous outcomes, and logistic regressions for effects on variances of
categorical outcomes.
Research Question Ten was investigated using an ANOVA of university sense of community for differences between types of residence, followed by linear regression analysis.