Data science education through education data: an end-to-end perspective

The increasing importance of data science has created a pressing need for more trained professionals. Hence, it is necessary to design efficient education techniques to rapidly train the next generation of students. We describe a holistic approach that integrates multiple objectives within a single semester course taught at the first year level. The objectives include basic data processing techniques, statistical analysis, teamwork, and project management. We conducted a project that combined different public datasets related to the cost and quality of education in the US. Students demonstrated a high degree of involvement in this exercise as they were curious to explore data that related to their own college selection process. It is well known that it is difficult to perform a cost-quality analysis in the field of college education. We used college ranking data created by the Wall Street Journal, and combined it with data from the Integrated Postsecondary Education Data System database (https://nces.ed.gov/ipeds/) that provided college costs and graduation data. Students worked in teams to download sections of the data from multiple sources, performed cleaning and verification, and merged the data for analysis. This taught them the values of teamwork, goal-setting, and process measurement for timely project completion. We determined that the Pearson correlation between the rank of a college and in-state tuition was -0.65 (p<; 0.005) for private colleges, and -0.34 (p<; 0.005) for public colleges. This indicates that higher ranked colleges tend to charge higher tuition. The Pearson correlation between the rank of all colleges and their graduation rates is -0.75 (p<; 0.005), and between rank and retention rates is -0.71 (p<; 0.005). Private college tuition is typically $40,000 annually, whereas public colleges cost $20,000. These figures suggest that public colleges provide good value. This has policy implications for government funding of education.