The Gaussian Graphical Model in Cross-Sectional and Time-Series Data.

Area of Science:

• Psychology
• Network Science
• Statistical Modeling

Background:

• Gaussian graphical models (GGMs) represent partial correlation networks, useful for exploratory data analysis.
• GGMs identify predictive relationships and model sparse covariance structures.
• They can suggest potential causal links between observed variables.

Purpose of the Study:

• To detail the utility of Gaussian graphical models (GGMs) in psychological research.
• To demonstrate GGM application across cross-sectional, single time-series, and multiple time-series datasets.
• To introduce graphical vector-autoregression (VAR) for time-series analysis and between-subjects networks.

Main Methods:

• Application of GGMs to independent cases (cross-sectional data).
• Utilizing GGMs within vector-autoregression (VAR) analysis for time-series data (graphical VAR).
• Constructing temporal, contemporaneous, and between-subjects networks from time-series and multi-subject data.

Main Results:

• GGMs effectively model covariance structures and identify inter-variable predictions in diverse psychological datasets.
• Graphical VAR yields interpretable temporal and contemporaneous networks for time-series analysis.
• Between-subjects networks can be derived from the covariance structure of stationary means.

Conclusions:

• Gaussian graphical models offer a powerful framework for network analysis in psychology.
• The proposed methods, implemented in R packages graphicalVAR and mlVAR, facilitate GGM estimation and interpretation.
• These network approaches enhance understanding of complex relationships in psychological data.