Group Integrative Dynamic Factor Models With Application to Multiple Subject Brain Connectivity.

Area of Science:

• Neuroscience
• Biostatistics
• Data Science

Background:

• Analyzing multi-subject time-series data requires methods that capture both inter-subject variability and intra-subject dynamics.
• Existing frameworks may not adequately address the complexities of group comparisons and longitudinal changes within individuals.

Purpose of the Study:

• To introduce a novel framework, GRoup Integrative DYnamic factor (GRIDY) models, for group-level analysis of multiple subjects' time-series data.
• To identify and characterize inter-subject similarities/differences between groups and intrasubject similarities/differences over time.
• To enable flexible reconstruction of latent factor series with adaptable covariance structures.

Main Methods:

• Development of the GRoup Integrative DYnamic factor (GRIDY) models.
• Integration of group spatial information with individual temporal dynamics.
• Application of a principal angle-based rank selection algorithm and a noniterative integrative analysis framework.
• Reconstruction of identifiable latent factor series inspired by simultaneous component analysis.

Main Results:

• Simulations demonstrated the robust performance of GRIDY models across various scenarios.
• The framework successfully identified inter- and intrasubject variations in time-series data.
• Application to resting-state fMRI data from autism spectrum disorder and control groups showcased its comparative capabilities.

Conclusions:

• The GRIDY models offer a powerful and flexible approach for group-level dynamic factor analysis.
• This framework enhances the understanding of inter-subject and intra-subject variability in complex datasets.
• GRIDY models show promise for neuroimaging research, particularly in comparing clinical and control groups.