Chapter 3: Research methods
3.8 Data analysis procedures
Data collected at the four time points were analysed using the Statistical Package for the Social Sciences (SPSS) version 16. Statistical significance was accepted at p ≤ 0.05. Descriptive and inferential statistics were used to summarise the data and address the research questions. The application of the different tests is shown below.
3.8.1 Descriptive statistics
Descriptive statistics were used to summarise the study population derived from the CSXQ and the CSEQ and time point one and time point two, respectively. Categorical data (for example, age, sex, ethnicity, place of residence and family academic history) were summarised using frequency counts and percentages.
69
Participant responses to questionnaire items and sub-scale scores were summarised using frequencies, mean, median, standard deviation and range as appropriate. Data collected at time points three and four were summarised using frequency counts and percentages.
3.8.2 Inferential statistics
Inferential statistical tests were used to draw conclusions from the data and address the research questions. Non-parametric tests were applied to the non-random sample and many variables not being normally distributed. This analysis was guided by a set of null hypotheses detailed below. Categorical data were transferred into cross tabulation Tables and chi-square analysis was used to test for association between appropriate variables (Greasley, 2008). In this study, the Wilcoxon signed rank test was used to test for differences between sub-scale scores for the CSXQ and the CSEQ. The Mann-Whitney U test is applied to test for difference in a continuous score between two independent groups (Pallant, 2010). In this study, the Mann-Whitney U was used to test for differences in variables between students who stayed and students who left in the first year of study. To assess whether or not the total score for ‘estimates of gains’ derived from the CSEQ was associated with final awards, the Kruskal-Wallis test was applied. The Kruskal-Wallis test is a non- parametric test for differences between three or more independent groups. The groupings were those students who achieved class one, class two or class three honours degrees. The Kruskal-Wallis test is a statistical technique that estimates the significance of differences between a set of means (Tabachnick & Fidell, 2007, Pallant, 2010).
The null hypotheses were:
H01: that there is no difference between the expectations and experiences of nursing students in their first year of university study.
The Wilcoxon Signed Rank test was used to test for expectations and experiences for the following sub-scales library use, relationships with other students and staff, academic effort, university environment.
70
H02: that there is no difference in characteristics between students who stay and students who leave.
The Chi- square statistic was used to test for associations between the study variables including age group, living and financial arrangements and parents’ academic history of students who stayed and students who left.
H03: that there was no difference in the level of expected academic effort for students who stayed and students who left in the first year of study.
The Mann-Whitney U was used to test for differences between students who stayed and students who left in sub-scale total scores for library use (3 items) learning (9 items); writing (4 items); university environment (7 items); general participation (6 items); use of computers (3 items).
H04: That there is no difference in expected social integration between students who stay and students who leave.
The Mann-Whitney U test was used to test for differences between students who stayed and students who left. In relation to scale scores for campus facilities (7 items); clubs and organisations (4 items); student acquaintances (7 items); academic effort (31 items); relationships with staff and students (2 items).
H05:that there is no association between the level of estimated gains and the final grade awarded.
The Kruskal-Wallis test was used to test for an association between the total score for estimated gains and the final grade awarded.
Model testing for students who stay and students who leave.
Logistic regression was used to predict the discrete outcomes ’stay’ or ’leave’. The predictor variables were age, educational qualifications, academic effort, interaction
71
with staff, campus activities and university environment (Tabachnick & Fidell, 2005). Logistic regression enabled assessment of how well the predictor variables explained the categorical dependent variable (Pallant, 2010). Probabilities were used to assess how much each variable contributed to the odds of belonging to the categories of ’stay’ or ’leave’, probability level was set at 0.05 (Sapsford, 2002). The logistic regression model also presents the chi-square value, degrees of freedom and the N value. The result of the predictors and odds ratio will be presented in chapter four.
3.8.3 Analysis of institutional data (individual and cohort completion)
The school of nursing and midwifery collects information routinely for all students on completion or exit from their programme of study. For this study, the required data was held on the school database and in the student records. Student records and the institutional tracker database were accessed to ascertain why students left the programme. The majority of students who left took an interruption in their studies with a view to resumption at a later date; therefore they did not complete exit questionnaires. The records of degree outcomes for the successful students of the September 2004 cohort were accessed and individual students were matched with their first year data by means of their roll numbers.