Modeling Cyclic Patterns Using a Two-Stage Hybrid Bayesian Approach

• In Event: AERA Roundtable Session Friday 7:45 am - Gold 4
  In Roundtable Session: RT4: Modeling Complex Data and Patterns: Bayesian, Longitudinal, and Machine Learning Approaches

Abstract

Cyclical phenomena are commonly observed, particularly in intensive longitudinal data. The conventional approach to analyzing cyclic patterns using the cosine function often suffers from a multiple-solution problem. To address this, researchers have reformulated the cosine function as a combination of sine and cosine terms. While this reformulation simplifies computation and resolves the multiple-solution issue, it complicates parameter interpretation and makes it difficult to assess how individual predictors influence features of the cyclic pattern. To bridge this gap, we propose a two-stage hybrid Bayesian approach that directly models cyclic pattern features while enabling evaluation of individual predictor effects. Through simulation studies, we demonstrate that the proposed method yields negligible bias and acceptable coverage rates.

Authors

• Han Du, University of California - Los Angeles

• Brian T. Keller, University of Missouri

• Lijuan Wang, University of Notre Dame