Stop-out is not attrition.
A computational framework for distinguishing temporary pauses from permanent withdrawal in self-paced online learning. Two documents take you from theory to practice — conceptual foundations, then a hands-on reference for applying the PSF library to real institutional data.
Can stop-out and attrition be distinguished computationally?
Two documents, one framework.
Begin with the tutorial to build understanding, then use the guide as a reference while working with data.
The Tutorial
From theory to practice
A guided walk through the Persistence-State Framework — from the problem of conflating stop-out with attrition, through the five-state model and latent Intent-to-Return, to a complete worked example on synthetic data.
- Five-state persistence model
- Latent Intent-to-Return variable
- HMM, DBN, survival, particle filter
- Worked example and 10 exercises
The User Guide
Reference for real data
A step-by-step reference for applying the PSF library to institutional data — collecting streams, preparing observation sequences, generating operational labels, fitting with multi-restart EM, and reading the diagnostic figures.
- Data collection and preparation
- Operational label generation
- Multi-restart EM fitting
- Matched-rate evaluation
The five states, at a glance.
Every student trajectory is a walk through this space. The critical edge is the one out of Stopped-Out.