Nonparametric Survival Analysis of the Loss Rate of Undergraduate Engineering Students
Article first published online: 2 JAN 2013
2011 American Society for Engineering Education
Journal of Engineering Education
Volume 100, Issue 2, pages 349–373, April 2011
How to Cite
Min, Y., Zhang, G., Long, R. A., Anderson, T. J. and Ohland, M. W. (2011), Nonparametric Survival Analysis of the Loss Rate of Undergraduate Engineering Students. Journal of Engineering Education, 100: 349–373. doi: 10.1002/j.2168-9830.2011.tb00017.x
- Issue published online: 2 JAN 2013
- Article first published online: 2 JAN 2013
- longitudinal study;
- nonparametric survival analysis;
As presented by Willet and Singer (1991), survival analysis can sensitively reveal rich information about when students leave their majors. Although survival analysis has been used to investigate student and faculty retention, it has not been applied to undergraduate engineering student retention.
The impact of cohort, gender, ethnicity, and SAT math and verbal scores on the loss rate of undergraduate engineering students was investigated to answer the questions: Does the profile of risk of dropout differ among groups with different backgrounds? When are students most likely to leave engineering? Which SAT scores better predict the risk of dropout?
Using a large longitudinal database that includes 100,179 engineering students from nine universities and spans 19 years, nonparametric survival analysis was adopted to obtain nonparametric estimates of survival and associated hazard functions, and complete rank tests for the association of variables.
There are significant differences for early semesters: White or female students tend to leave engineering earlier than other populations. Engineering students leave engineering during the third semester the most, although students who have an SAT math score less than 550 tend to leave engineering during the second semester. SAT math score better predicts the risk of dropout than SAT verbal score.
The results of this study support using survival analysis to better understand factors that determine student success since student retention is a dynamic problem. Survival analysis allows characteristics such as risk to be evaluated by semester, giving insight to when interventions might be most effective.