Modeling the time to dropout under phase-wise variable stress fixed cohort setup

The event of a student dropping out from an academic program depends on several factors namely course content, change in interest, financial problems among many others. These factors vary interdependently with different phases of the academic program. We assume that the factors put different amount of academic stresses on a student in different phases of the program. We formulate and analyze such an accumulated-stress model under the assumption that the attrition time at each phase follows the Kumaraswamy distribution. A hazard-rate based approach is used to model the accumulated stress through different phases. At each phase the stress levels vary. Accordingly, we estimate the model parameters based on the frequentist approach. Extensive simulation experiments indicate satisfactory performance of the estimators. A synthetic dataset related to students' dropout has been analyzed for illustrative purpose.