A Nonlinear State Space Model Predicting Dropout: The Case of Special Education Students in the Hellenic Open University

While open and distance education gains growing recognition over time, it also faces increasing drop-out rates. Consequently, the development of predictive models for early identification of students at-risk for drop-out could be critical to promote ongoing engagement. This study aims to gain insights into the dropout prediction problem in a sample of postgraduate special education students. Therefore, a nonlinear state-space model was employed. An Expectation-Maximization (EM) algorithm that iterates between state estimation (E-step) and parameter estimation (M-step) was retrieved from the existing literature. However, the variables were redefined according to available data. In the dataset a total number of n[subscript 1] = 1337 students were enrolled, attending n[subscript 2] = 7 different modules. Statistical analysis showed that the majority of the students dropped out during the second week. Moreover, an inversely proportional relationship was observed between the dropout rates and the number of weeks that the student has actively engaged in the module. Significant differences across modules were also observed. Results are discussed in terms of their application in the training and education of the collaborating teaching staff working in the HOU. Future work in the field should be expanded in order to investigate students' drop-out in more courses other than special education.