Evaluating Covariate Effects on ESM Measurement Model Changes with Latent Markov Factor Analysis: A Three-Step...

Area of Science:

• Psychometrics
• Longitudinal Data Analysis
• Statistical Modeling

Background:

• Measurement model (MM) invariance is crucial for valid inferences in intensive longitudinal data.
• Latent Markov factor analysis (LMFA) was proposed to evaluate MM invariance over time.
• Existing one-step LMFA estimation is complex and hinders covariate analysis.

Purpose of the Study:

• To simplify the estimation of latent Markov factor analysis (LMFA).
• To facilitate the exploration of covariate effects on state memberships in LMFA.
• To provide an intuitive and practical approach for applied researchers.

Main Methods:

• A three-step estimation procedure for LMFA is introduced.
• Step 1: Mixture factor analysis to obtain states (treating repeated measures as independent).
• Step 2: Assign observations to states.
• Step 3: Latent Markov modeling incorporating classification errors.

Main Results:

• The proposed method simplifies LMFA estimation compared to the one-step FIML approach.
• The three-step method allows for intuitive exploration of covariate effects on state memberships.
• A real data example demonstrates the empirical utility and value of the approach.

Conclusions:

• The simplified three-step LMFA estimation enhances practical application in psychological research.
• This method improves the accessibility of analyzing dynamic psychological factors using intensive longitudinal data.
• It offers a more manageable approach for model selection and covariate analysis in LMFA.