Predicting Post Secondary Educational Outcomes With Survival Analysis

Identifying potential students and understanding what affects their decision to depart the track of obtaining a college degree in engineering is critical to engineering educational research. This study used data from the National Education Longitudinal Study of 1988-2000 (NELS) to develop a model for predicting post-secondary educational outcomes with particular focus on students earning a college degree in Science, Technology, Engineering, or Mathematics (STEM). The objective was to identify factors that affected the probability of a given student “surviving” on the STEM track past a key time point in the study at which most students attending college were nearing graduation. NELS provided a comprehensive set of variables for 12,144 students of which 11,128 had clearly determinable educational outcomes. The set of potential outcomes included dropping out of high school, completing education at high school graduation, dropping out of college, earning a less than four year degree, earning a degree other than STEM, earning a STEM degree, or having an incomplete degree at the study’s conclusion. The modeling process utilized demographic, attitudinal, and academic performance data mainly collected at the 8 th grade level as well as standardized test scores and college enrollment status variables. Survival analysis models were fit with randomly selected data and then applied to reserved test data to determine the models’ sensitivity and specificity in predicting a STEM vs. other outcome. The models performed well in distinguishing between STEM students and those who did not successfully complete a college degree. The different categories of educational outcomes exhibited markedly different hazard curves showing the periods of greatest STEM track departure risk varied between student outcomes. This suggested that the optimum times at which to offer positive interventions to keep students on the potential STEM track vary by the type of outcome they are otherwise likely to experience. This includes encouraging students to remain in high school, to apply to college, and to persist once enrolled in an STEM program.