oWS6: Just in Time or Just a Guess? Validating Prediction Models Based on Longitudinal Data
This workshop is designed for researchers working with longitudinal data to develop predictive models using machine learning. Longitudinal data involve repeated measurements from the same individuals, such as daily surveys or passively collected data via smartphones and wearable devices.
Participants will explore key challenges in evaluating prediction models, including selecting appropriate training and test sets. The workshop will review common use case scenarios and validation methods, demonstrating their alignment with different research goals. Focus will be placed on potential pitfalls when evaluating models and recommendations to avoid them.
Attendees will also be introduced to a tool (https://github.com/AnnaLangener/Justintime) designed to help researchers assess misalignments between validation strategies and research objectives.