Forecasting Intra-individual Changes of Affective States Taking into Account Inter-individual Differences Using...

Area of Science:

• Psychology
• Educational Psychology
• Statistics

Background:

• University student dropout in STEM is a complex longitudinal process influenced by individual differences and changes.
• Existing dynamic latent variable models (e.g., DSEM) partially address complex data structures in longitudinal studies.
• Forecasting in dynamic frameworks for real-time inference remains an under-researched area, particularly for student dropout.

Purpose of the Study:

• To develop and demonstrate a Bayesian forecasting method for multivariate intra-individual variables within dynamic latent variable frameworks.
• To apply this forecasting method to intensive longitudinal data (ILD) from a math dropout study.
• To enable real-time prediction of critical states related to student dropout for timely intervention.

Main Methods:

• Utilized a Forward Filtering Backward Sampling method for Bayesian forecasting.
• Applied the method to dynamic latent variable models (DLVMs) for intensive longitudinal data (ILD).
• Analyzed data from a large university student dropout study in math with 50 measurement occasions.

Main Results:

• Successfully forecasted multivariate intra-individual variables and time-dependent class membership (affective states).
• Demonstrated the ability to predict critical dynamic states, such as stress or pre-decisional states, up to 8 weeks before dropout.
• Integrated ILD, dynamic frameworks, and forecasting for empirical study on math student dropout.

Conclusions:

• The proposed Bayesian forecasting method effectively predicts critical states in dynamic frameworks using ILD.
• This approach allows for early identification of students at risk of dropout, facilitating proactive interventions.
• The study provides a practical tool for understanding and mitigating student attrition in STEM fields.