Projecting Student Trajectories in a Computer-Assisted Instruction Curriculum

Ten models are proposed for predicting a student’s final grade placement in a computer-assisted instruction curriculum from the time the student spends taking lessons. Using data from approximately 2,000 elementary students, the models are tested both for ability to predict final grade placement and for ability to describe all points observed throughout the year. Two of the simplest models, using only the most recent point and parameters estimated for the whole group, are best at prediction. The power function model with all parameters estimated individually for each student, while relatively poor at prediction, is best for describing all observed points. Several methods forpredicting the amqunt of time a student will spend taking lessons during a year are also tested.