A Latent Gaussian process model for analysing intensive longitudinal data.

Area of Science:

• Psychology
• Social Sciences
• Data Science

Background:

• Intensive longitudinal studies are increasingly common in psychology, enabled by new technologies for high-temporal-resolution data collection.
• Existing latent curve models face limitations with the unique characteristics of intensive longitudinal data, such as dense and unequally spaced observations.

Purpose of the Study:

• To introduce a novel modeling framework for latent curve analysis tailored for intensive longitudinal data.
• To better capture individual-specific continuous-time latent processes and address challenges of time-varying data.

Main Methods:

• A Gaussian process model is employed to represent the individual-specific continuous-time latent process.
• The framework is presented as a semi-parametric extension of traditional latent curve models within structural equation modeling.
• Parameter estimation and statistical inference are conducted using an empirical Bayes approach.

Main Results:

• The proposed Gaussian process latent curve model effectively handles intensive longitudinal data, including irregularly timed observations.
• Simulation studies demonstrate the validity and performance of the new modeling framework.
• The model's utility is illustrated through an ecological momentary assessment data analysis.

Conclusions:

• The novel Gaussian process-based latent curve analysis framework offers a more suitable approach for intensive longitudinal data in psychology.
• This method enhances the understanding of individual dynamics by modeling continuous-time latent processes.
• The framework provides a robust tool for analyzing complex, time-varying psychological data.