Modeling cyclic patterns using a two-stage hybrid Bayesian approach

Area of Science:

• Statistics
• Data Analysis
• Bayesian Methods

Background:

• Cyclical phenomena are prevalent in intensive longitudinal data.
• Traditional cosine function analysis for cyclic patterns faces a multiple-solution problem.
• Reformulating cosine functions simplifies computation but hinders parameter interpretation.

Purpose of the Study:

• To propose a novel two-stage hybrid Bayesian approach for analyzing cyclic patterns.
• To enable direct modeling of cyclic pattern features.
• To allow for the evaluation of individual predictor effects on cyclic patterns.

Main Methods:

• A two-stage hybrid Bayesian framework is introduced.
• The approach directly models features of cyclic patterns.
• Simulation studies are used to validate the method's performance.

Main Results:

• The proposed Bayesian method demonstrates negligible bias in parameter estimation.
• Acceptable coverage rates were achieved in simulation studies.
• The approach effectively models cyclic pattern features and predictor influences.

Conclusions:

• The novel Bayesian approach effectively addresses limitations in analyzing cyclic patterns.
• It provides a robust method for understanding how predictors influence cyclic features in longitudinal data.
• This method offers enhanced interpretability and accuracy for cyclical data analysis.