To explore the relationship between curricular complexity and study abroad participation across 48 majors at Purdue University. To accomplish this purpose, we will address the following research questions: RQ1. What are the curriculum complexity and study abroad participation rates for the largest majors at Purdue University? RQ2. What is the correlation between curriculum complexity and study abroad participation at Purdue University? RQ3. Is there a significant difference in curricular complexity or study abroad participation across colleges at Purdue University?
Our project can provide insights to engineering programs seeking to improve study abroad participation about the challenges that may arise from curricular complexity and what strategies may help address this issue.
To address RQ1, we generated summary statistics for curricular complexity and study abroad participation rates across the 48 majors in our data set. Then, to address RQ2, we created a correlation matrix to identify the relationships between the variables shown in Table 3. We used Pearson’s correlation with a Holm correction for the correlation matrix since all our variables are continuous (Field et al., 2012) and Pearson’s correlation is robust even in cases where the data is nonnormal (Norman, 2010). Lastly, to address RQ3, we ran the Kruskal-Wallis rank sum test to find out whether there are differences across colleges in curricular complexity or total study abroad participation rates. We chose the Kruskal-Wallis test because our variables violated the assumptions of normality and homoscedasticity and we have small sample sizes for each college (i.e., less than ten majors per college). The Kruskal-Wallis test is a nonparametric alternative to ANOVA that allows for a comparison of means across groups (Field et al., 2012). We only included colleges that had at least five majors in the data set because the p-values can be inaccurate if there are fewer than five observations in each group (Field et al., 2012). Therefore, only six out of ten colleges were included in the analysis for RQ3. We used the Wilcoxon rank sum test with the Benjamini-Hochberg correction to conduct post-hoc tests to do pairwise comparisons between groups.