This third article in a series describing survival analysis of engineering student retention and graduation introduces accelerated failure-time as an alternative to the Cox proportional hazards model to the context of student data. The new survival analysis of graduation data presented here assumes different distributions including exponential, lognormal and Weibull, and assesses efficiency and goodness of fit based on estimated parameters, likelihood and number of observations. Results are associated with the effects of American College Test and Scholastic Assessment Test scores, gender, and other demographic information on retention and graduation. Some results confirm what we have previously learned from proportional hazards models of graduation, and some results are unique to accelerated failure-time models. (Contains 7 tables.)